Загрузка видео...

Не удалось загрузить видео

На главную

As promised, instead of for $15,000 GPUs, we max optimized a model that can fit comfortably in any GPU with 8GB VRAM running Windows 200 tok/s, *128k context, text/image/audio. Anyone who wants to join the local AI community with 5 year old laptop can run this with recipe below,...

net_termina's profile picture

Killy

3,441 subscribers

125,996 просмотров • 8 дней назад •via X (Twitter)

Наука и технологииОбразование
LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

Run Gemma 4 26B MoE on 8GB VRAM with 250k context at 20+ tokens/sec If you own any 8GB VRAM graphics card, stop what you are doing. Local AI just had its absolute "Holy Shit" moment for budget hardware. Yesterday, I benchmarked Unsloth Gemma 4 12B Q4_K_XL on an 8GB card. The community went wild but immediately demanded more: "Can we run a 25B+ model on budget GPUs?" Today, I’m delivering exactly that. I am running a massive 26B parameter Mixture of Experts (MoE) model locally on a standard 8GB VRAM setup with 250k full native context!. If you own an RTX 3060, 3070, 4060, or any budget GPU with 8GB of VRAM, the local AI paradigm has completely changed. The performance metrics are astonishing: - 20 tokens/sec flat decode throughput. - Stable, flat decode speed even with massive prompts. - I threw a 60k token prompt at it, and it still clocked in at 20 TPS without dropping a single frame. # What about prefill? Yes, Time To First Token (TTFT) is slightly high when swallowing massive contexts. But with a solid 200 tokens/sec prefill speed, the wait is barely noticeable and highly usable. And this is running completely without Multi Token Prediction (MTP) active. How is this possible? It’s the magic of Google's new QAT (Quantization Aware Training) quants for Gemma 4. The model weight file (unsloth gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf) is only 13.2 GB, making it the ultimate local powerhouse. # The Test Setup: CPU: Intel Core i7 RAM: 16GB System RAM GPU: NVIDIA GeForce RTX 4060 Laptop GPU (8GB VRAM) # The Secret Sauce (The -cmoe Flag) To make this work properly on any 8GB card, you must use the -cmoe (CPU MoE) flag in llama.cpp. This flag isolates the heavy MoE expert weights directly to system memory (CPU/RAM) while letting your GPU focus strictly on the Attention layers and the KV Cache. It prevents VRAM spillage and holds the throughput rock solid. # The flags: -m "gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf" -cmoe -c 248000 -v Once running, just open the UI on localhost and toggle the new reasoning lightbulb icon in the text input box to watch the model perform multi step thinking. Are you still running smaller models, or are you ready to scale up your budget local setups? Let's discuss in the replies
0:19

Run Gemma 4 26B MoE on 8GB VRAM with 250k context at 20+ tokens/sec If you own any 8GB VRAM graphics card, stop what you are doing. Local AI just had its absolute "Holy Shit" moment for budget hardware. Yesterday, I benchmarked Unsloth Gemma 4 12B Q4_K_XL on an 8GB card. The community went wild but immediately demanded more: "Can we run a 25B+ model on budget GPUs?" Today, I’m delivering exactly that. I am running a massive 26B parameter Mixture of Experts (MoE) model locally on a standard 8GB VRAM setup with 250k full native context!. If you own an RTX 3060, 3070, 4060, or any budget GPU with 8GB of VRAM, the local AI paradigm has completely changed. The performance metrics are astonishing: - 20 tokens/sec flat decode throughput. - Stable, flat decode speed even with massive prompts. - I threw a 60k token prompt at it, and it still clocked in at 20 TPS without dropping a single frame. # What about prefill? Yes, Time To First Token (TTFT) is slightly high when swallowing massive contexts. But with a solid 200 tokens/sec prefill speed, the wait is barely noticeable and highly usable. And this is running completely without Multi Token Prediction (MTP) active. How is this possible? It’s the magic of Google's new QAT (Quantization Aware Training) quants for Gemma 4. The model weight file (unsloth gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf) is only 13.2 GB, making it the ultimate local powerhouse. # The Test Setup: CPU: Intel Core i7 RAM: 16GB System RAM GPU: NVIDIA GeForce RTX 4060 Laptop GPU (8GB VRAM) # The Secret Sauce (The -cmoe Flag) To make this work properly on any 8GB card, you must use the -cmoe (CPU MoE) flag in llama.cpp. This flag isolates the heavy MoE expert weights directly to system memory (CPU/RAM) while letting your GPU focus strictly on the Attention layers and the KV Cache. It prevents VRAM spillage and holds the throughput rock solid. # The flags: -m "gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf" -cmoe -c 248000 -v Once running, just open the UI on localhost and toggle the new reasoning lightbulb icon in the text input box to watch the model perform multi step thinking. Are you still running smaller models, or are you ready to scale up your budget local setups? Let's discuss in the replies

Alok

292,770 просмотров • 3 месяцев назад

GEMMA 4 26B ON AN RTX 4060 WITH A 248K TOKEN CONTEXT WINDOW 20 tokens per second and a context window so large you can feed it entire codebases, books and research papers in a single prompt this is not a cloud api and not a server rack, this is a regular consumer gpu running locally with llama.cpp and q4_k_xl quantization 248k context on an 8gb vram card was not supposed to be possible and here it is just running on someone’s desk the article below covers exactly which tools and configs make this kind of setup work in 2026 ↓
0:19

GEMMA 4 26B ON AN RTX 4060 WITH A 248K TOKEN CONTEXT WINDOW 20 tokens per second and a context window so large you can feed it entire codebases, books and research papers in a single prompt this is not a cloud api and not a server rack, this is a regular consumer gpu running locally with llama.cpp and q4_k_xl quantization 248k context on an 8gb vram card was not supposed to be possible and here it is just running on someone’s desk the article below covers exactly which tools and configs make this kind of setup work in 2026 ↓

leopardracer

56,865 просмотров • 3 месяцев назад

Gemma 4 12B QAT (dense) achieves 1000+ tokens/sec prefill on 8GB VRAM with 120k context Gemma 4 12B QAT (dense), TurboQuant (Without MTP), RTX 4060 8GB VRAM: Prefill: 1000+ tok/s (42% increase) Decode: 25+ tok/s (25% increase) Context: 120k (150% increase) prefill was 700 tok/sec and decode 20 tok/sec with only 48k context without turbo quant (older test with mtp link in the comments) llama.cpp TurboQuant flags: -m gemma-4-12B-it-qat-UD-Q4_K_XL.gguf -c 120000 --cache-type-k q8_0 --cache-type-v turbo3 -ngl 99 --port 8080 tested with a 27k prompt, 120k context loaded. -ngl 99 here isn't a typo, full 12B dense, every layer on GPU, on an 8GB card. that's the part worth sitting with. The model has vision, audio input, thinking/reasoning and fits your 8GB card. TurboQuant's KV cache savings are what free up the room to do that at 120k context. side by side with yesterday: 26B A4B MoE got 320+ tok/s prefill. this dense 12B is clearing 1000+ rig: RTX 4060 8GB · i7H · 16GB RAM same two flags as yesterday, different model size: --cache-type-k q8_0 --cache-type-v turbo3 thanks to TheTom/llama-cpp-turboquant, TurboQuant fork of llama.cpp by Tom Turney (Tom Turney) to make this work. unsloth's model quant huggingface and the llama.cpp fork github link in the comments Do you prefer a dense or a MoE for your 8GB card?
0:12

Gemma 4 12B QAT (dense) achieves 1000+ tokens/sec prefill on 8GB VRAM with 120k context Gemma 4 12B QAT (dense), TurboQuant (Without MTP), RTX 4060 8GB VRAM: Prefill: 1000+ tok/s (42% increase) Decode: 25+ tok/s (25% increase) Context: 120k (150% increase) prefill was 700 tok/sec and decode 20 tok/sec with only 48k context without turbo quant (older test with mtp link in the comments) llama.cpp TurboQuant flags: -m gemma-4-12B-it-qat-UD-Q4_K_XL.gguf -c 120000 --cache-type-k q8_0 --cache-type-v turbo3 -ngl 99 --port 8080 tested with a 27k prompt, 120k context loaded. -ngl 99 here isn't a typo, full 12B dense, every layer on GPU, on an 8GB card. that's the part worth sitting with. The model has vision, audio input, thinking/reasoning and fits your 8GB card. TurboQuant's KV cache savings are what free up the room to do that at 120k context. side by side with yesterday: 26B A4B MoE got 320+ tok/s prefill. this dense 12B is clearing 1000+ rig: RTX 4060 8GB · i7H · 16GB RAM same two flags as yesterday, different model size: --cache-type-k q8_0 --cache-type-v turbo3 thanks to TheTom/llama-cpp-turboquant, TurboQuant fork of llama.cpp by Tom Turney (Tom Turney) to make this work. unsloth's model quant huggingface and the llama.cpp fork github link in the comments Do you prefer a dense or a MoE for your 8GB card?

Alok

34,500 просмотров • 2 месяцев назад

my 8 GB VRAM gaming laptop is absolutely going to hate me for this. but I still did it. ran a 31b dense model (Gemma 4 31b Q4) with only 8 GB VRAM last week I ran Gemma 4 26B A4B a mixture of experts model on my RTX 4060 and hit 25–28 tokens/sec using llama.cpp's new MTP support. smooth. snappy. but MoE has a secret: it only activates 4B parameters per token despite having 26B total. that's why it flies. so the real question started haunting me. what if I throw a full, no tricks, every parameter fires on every token, 31B DENSE model at the same machine? # Hardware: GPU: NVIDIA RTX 4060, 8 GB VRAM RAM: 16 GB CPU: Intel Core i7 H Laptop. Gaming. Modest. The model: gemma-4-31B-it-qat-UD-Q4_K_XL.gguf (model's unsloth huggingface link in the comments) This is Google DeepMind's flagship dense model in the Gemma 4 family that can run on single consumer GPU. It packs a hybrid attention architecture, supports up to 256K context natively, and is QAT (Quantization Aware Training) optimized, meaning it retains far more quality than standard post training quants at the same bit depth. This is NOT the MoE. This is 31 BILLION dense parameters, every single one of them loaded. # the flags I used: -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -cnv --spec-type draft-mtp --spec-draft-model mtp-gemma-4-31B-it.gguf --spec-draft-n-max 8 --spec-draft-p-min 0.6 -c 6000 -v Multi Token Prediction (MTP) is still active here. Separate draft GGUF required, same as the 26B setup. # Results: → Decode: ~3 tokens/sec → Prefill: ~2 tokens/sec → Context: 6000 tokens → Hardware crying quietly in the corner: yes so is 3 tps actually usable? For real time back and forth chat? Not ideal. You're not having a fluid conversation at 3 tps. but slow ≠ useless. And this is where it gets genuinely interesting. think about how senior devs actually work in a real team. But when something is architectural, deeply complex, or needs serious reasoning? they walk down the hall and escalate to the senior. That's exactly the local AI agent architecture this unlocks: → Fast orchestrator model (Gemma 4 26B MoE at 25+ tps) handles routing, simple queries, tool calls, memory. The junior dev. → Gemma 4 31B dense is the senior, called only when the fast model genuinely hits a wall. Hard multi step reasoning. Complex code generation. Deep architectural decisions. The agentic loop stays fast. Only the hard hops touch the 31B. That's a legitimate production grade local AI architecture on a budget hardware. (requires 2 8gb gpus) other workflows where 3 tps is completely fine: - overnight batch jobs. summarize documents, extract structured data, review code. Fire it off. Sleep. wake up to results. - One shot deep reasoning - Silent code audit loops, you write and test, the 31B reviews diffs and flags issues in the background between your sprints - Any workflow where output quality > output speed A few weeks ago, nobody was running a 30B+ dense model on a single consumer GPU with 8 GB VRAM. At all. Now we're doing it on an Intel i7-H gaming laptop with a NVIDIA RTX 4060, thanks to llama.cpp + QAT quants + MTP speculative drafting. Google DeepMind said the Gemma 4 31B targets "consumer GPUs and workstations." They were not exaggerating. The hardware bar to run serious frontier class models locally keeps dropping. the tools are here. the models are here. you just have to be willing to abuse your laptop a little. what workflows would you actually run on a local 3 tps 31B dense model? genuinely curious. drop it below.
0:18

my 8 GB VRAM gaming laptop is absolutely going to hate me for this. but I still did it. ran a 31b dense model (Gemma 4 31b Q4) with only 8 GB VRAM last week I ran Gemma 4 26B A4B a mixture of experts model on my RTX 4060 and hit 25–28 tokens/sec using llama.cpp's new MTP support. smooth. snappy. but MoE has a secret: it only activates 4B parameters per token despite having 26B total. that's why it flies. so the real question started haunting me. what if I throw a full, no tricks, every parameter fires on every token, 31B DENSE model at the same machine? # Hardware: GPU: NVIDIA RTX 4060, 8 GB VRAM RAM: 16 GB CPU: Intel Core i7 H Laptop. Gaming. Modest. The model: gemma-4-31B-it-qat-UD-Q4_K_XL.gguf (model's unsloth huggingface link in the comments) This is Google DeepMind's flagship dense model in the Gemma 4 family that can run on single consumer GPU. It packs a hybrid attention architecture, supports up to 256K context natively, and is QAT (Quantization Aware Training) optimized, meaning it retains far more quality than standard post training quants at the same bit depth. This is NOT the MoE. This is 31 BILLION dense parameters, every single one of them loaded. # the flags I used: -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -cnv --spec-type draft-mtp --spec-draft-model mtp-gemma-4-31B-it.gguf --spec-draft-n-max 8 --spec-draft-p-min 0.6 -c 6000 -v Multi Token Prediction (MTP) is still active here. Separate draft GGUF required, same as the 26B setup. # Results: → Decode: ~3 tokens/sec → Prefill: ~2 tokens/sec → Context: 6000 tokens → Hardware crying quietly in the corner: yes so is 3 tps actually usable? For real time back and forth chat? Not ideal. You're not having a fluid conversation at 3 tps. but slow ≠ useless. And this is where it gets genuinely interesting. think about how senior devs actually work in a real team. But when something is architectural, deeply complex, or needs serious reasoning? they walk down the hall and escalate to the senior. That's exactly the local AI agent architecture this unlocks: → Fast orchestrator model (Gemma 4 26B MoE at 25+ tps) handles routing, simple queries, tool calls, memory. The junior dev. → Gemma 4 31B dense is the senior, called only when the fast model genuinely hits a wall. Hard multi step reasoning. Complex code generation. Deep architectural decisions. The agentic loop stays fast. Only the hard hops touch the 31B. That's a legitimate production grade local AI architecture on a budget hardware. (requires 2 8gb gpus) other workflows where 3 tps is completely fine: - overnight batch jobs. summarize documents, extract structured data, review code. Fire it off. Sleep. wake up to results. - One shot deep reasoning - Silent code audit loops, you write and test, the 31B reviews diffs and flags issues in the background between your sprints - Any workflow where output quality > output speed A few weeks ago, nobody was running a 30B+ dense model on a single consumer GPU with 8 GB VRAM. At all. Now we're doing it on an Intel i7-H gaming laptop with a NVIDIA RTX 4060, thanks to llama.cpp + QAT quants + MTP speculative drafting. Google DeepMind said the Gemma 4 31B targets "consumer GPUs and workstations." They were not exaggerating. The hardware bar to run serious frontier class models locally keeps dropping. the tools are here. the models are here. you just have to be willing to abuse your laptop a little. what workflows would you actually run on a local 3 tps 31B dense model? genuinely curious. drop it below.

Alok

63,689 просмотров • 2 месяцев назад

Run Gemma 4 26b MTP on 8 GB VRAM GPUs at 25+ tokens/second. Flags included! local llm space is moving at terminal velocity. only 3 days ago google released gemma 4 26b a4b qat quants. more efficient than before, ran on 8gb vram at 20 tok/sec. and now just a few hours ago, mainline llama.cpp merged a massive update and we just shattered our own record. decode throughput went 25-40% up on the same 8 GB VRAM setup! Before MTP: 20 tps -> After MTP: 28 tps! llama.cpp just officially merged PR #23398 ("add Gemma4 MTP"), bringing native Multi-Token Prediction (MTP) support to Gemma 4 models. By running speculative drafting on the same 8GB VRAM RTX 4060 setup, my decode throughput on a 64k context instantly leaped to a blistering 25–27 tokens/sec thats 25-30% increase with the same hardware. Here is the architectural catch you need to know: Unlike the Qwen 3.5 and 3.6 series, which bake the MTP heads directly into the base GGUF, the Gemma 4 MTP head is not built in. You must download a separate, specialized MTP drafter GGUF (the assistant model) to act as the speculator. (I've dropped the download link in the replies). copy and try the exact flags: -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --spec-type draft-mtp --spec-draft-n-max 6 --spec-draft-p-min 0.7 --spec-draft-model gemma-4-26b-A4B-it-assistant-Q4_0.gguf -c 64000 -v n-max 4 and p-min 0.7 is also worth checking out. benchmark on your setup and workflow. if you have a single 8 gb vram nvidia rtx 4060, 3060, 3070, 2080, 2070, grab the MTP drafter GGUF link in the comments and try it yourself. Check it out even if you have asmaller or a larger gpu, such as a single rtx 3090, 4090, 3060, 2060. MTP works for all gemma 4 sizes such as gemma 4 12b, gemma 4 31b etc. but remember to grab the correct mtp draft assistant models respectively. what are you benchmarking today
0:15

Run Gemma 4 26b MTP on 8 GB VRAM GPUs at 25+ tokens/second. Flags included! local llm space is moving at terminal velocity. only 3 days ago google released gemma 4 26b a4b qat quants. more efficient than before, ran on 8gb vram at 20 tok/sec. and now just a few hours ago, mainline llama.cpp merged a massive update and we just shattered our own record. decode throughput went 25-40% up on the same 8 GB VRAM setup! Before MTP: 20 tps -> After MTP: 28 tps! llama.cpp just officially merged PR #23398 ("add Gemma4 MTP"), bringing native Multi-Token Prediction (MTP) support to Gemma 4 models. By running speculative drafting on the same 8GB VRAM RTX 4060 setup, my decode throughput on a 64k context instantly leaped to a blistering 25–27 tokens/sec thats 25-30% increase with the same hardware. Here is the architectural catch you need to know: Unlike the Qwen 3.5 and 3.6 series, which bake the MTP heads directly into the base GGUF, the Gemma 4 MTP head is not built in. You must download a separate, specialized MTP drafter GGUF (the assistant model) to act as the speculator. (I've dropped the download link in the replies). copy and try the exact flags: -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --spec-type draft-mtp --spec-draft-n-max 6 --spec-draft-p-min 0.7 --spec-draft-model gemma-4-26b-A4B-it-assistant-Q4_0.gguf -c 64000 -v n-max 4 and p-min 0.7 is also worth checking out. benchmark on your setup and workflow. if you have a single 8 gb vram nvidia rtx 4060, 3060, 3070, 2080, 2070, grab the MTP drafter GGUF link in the comments and try it yourself. Check it out even if you have asmaller or a larger gpu, such as a single rtx 3090, 4090, 3060, 2060. MTP works for all gemma 4 sizes such as gemma 4 12b, gemma 4 31b etc. but remember to grab the correct mtp draft assistant models respectively. what are you benchmarking today

Alok

200,913 просмотров • 3 месяцев назад

I told you to claim your free 16GB NVIDIA GPU for learning Local LLMs. Now I’m going to show you how to double its inference speed without touching the hardware. Google Colab gives you an enterprise grade NVIDIA Tesla T4 GPU for free, roughly 4 hours every single day. It is the absolute perfect sandbox for learning AI engineering, testing inference flags, and pushing massive context windows. The local AI timeline is moving way too fast. If you aren't using Multi Token Prediction (MTP) yet, you are leaving massive performance on the table. I just pushed DeepMind’s Gemma 4 26B to 64.9 t/s on this exact free tier. Let's look at the raw benchmark data running on an Ubuntu Linux environment with the latest compiled llama.cpp binaries and quantized GGUFs from Unsloth via HuggingFace: # Qwen 3.5 9B (Dense): Base: [ Prompt: 626.7 t/s | Generation: 21.0 t/s ] With MTP: [ Prompt: 539.1 t/s | Generation: 24.8 t/s ] # Gemma 4 26B QAT (MoE): Base: [ Prompt: 634.2 t/s | Generation: 48.3 t/s ] With MTP: [ Prompt: 572.1 t/s | Generation: 64.9 t/s ] If you are paying attention, this single Colab notebook reveals 3 massive observations about the current state of local LLMs: # 1. The MTP Speedup (Software Overclocking) Standard autoregressive decoding guesses one token at a time. MTP acts like a highly optimized, built in speculative decoder. It predicts multiple future tokens at once and the main model verifies them in parallel. The result? Zero accuracy loss and a massive throughput increase. Gemma jumped from 48 to 65 t/s just by flipping a flag. # 2. The MoE Paradox (Bigger is Faster) How does a 26B parameter model absolutely destroy a 9B model in raw speed on the exact same hardware? Architecture. Qwen 3.5 9B is a dense model. it activates all 9 billion parameters for every single token. Gemma 4 26B is a Mixture of Experts (MoE) model. It routes data efficiently, activating only 4B parameters per token. You get the reasoning capabilities of a 26B model with the compute cost of a 4B model. 3. Thinking Efficiency When I ran the exact same complex prompt on both models, the larger MoE spent significantly fewer "thinking" tokens to arrive at the correct answer. A smarter model doesn't just give better answers; it gets to the point faster, saving you compute cycles and preserving your context window. # Want to run this yourself? Here are the exact llama.cpp CLI commands. For Qwen (MTP is baked into the main model): ./llama-cli -m Qwen3.5-9B-UD-Q4_K_XL.gguf -p "Explain quantum computing." -n 2000 -c 8000 -ngl 99 -fa on --spec-type draft-mtp --spec-draft-n-max 4 --spec-draft-p-min 0.7 For Gemma (Using a separate lightweight draft model): ./llama-cli -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --model-draft mtp-gemma-4-26B-A4B-it.gguf -p "Explain quantum computing." -n 2000 -c 8000 -ngl 99 -fa on --spec-type draft-mtp --spec-draft-n-max 4 --spec-draft-p-min 0.7 Stop waiting for a $3,000 rig. Boot up Colab, pull these models, and start building your stack. I’ve put together a completely free, cell by cell Google Colab notebook that automates this entire workflow so you can test it yourself in 5 minutes and learn. Link to the notebook is in the comments below. Experiemt with different MTP parameters, context windows and post your results in the comments.
0:12

I told you to claim your free 16GB NVIDIA GPU for learning Local LLMs. Now I’m going to show you how to double its inference speed without touching the hardware. Google Colab gives you an enterprise grade NVIDIA Tesla T4 GPU for free, roughly 4 hours every single day. It is the absolute perfect sandbox for learning AI engineering, testing inference flags, and pushing massive context windows. The local AI timeline is moving way too fast. If you aren't using Multi Token Prediction (MTP) yet, you are leaving massive performance on the table. I just pushed DeepMind’s Gemma 4 26B to 64.9 t/s on this exact free tier. Let's look at the raw benchmark data running on an Ubuntu Linux environment with the latest compiled llama.cpp binaries and quantized GGUFs from Unsloth via HuggingFace: # Qwen 3.5 9B (Dense): Base: [ Prompt: 626.7 t/s | Generation: 21.0 t/s ] With MTP: [ Prompt: 539.1 t/s | Generation: 24.8 t/s ] # Gemma 4 26B QAT (MoE): Base: [ Prompt: 634.2 t/s | Generation: 48.3 t/s ] With MTP: [ Prompt: 572.1 t/s | Generation: 64.9 t/s ] If you are paying attention, this single Colab notebook reveals 3 massive observations about the current state of local LLMs: # 1. The MTP Speedup (Software Overclocking) Standard autoregressive decoding guesses one token at a time. MTP acts like a highly optimized, built in speculative decoder. It predicts multiple future tokens at once and the main model verifies them in parallel. The result? Zero accuracy loss and a massive throughput increase. Gemma jumped from 48 to 65 t/s just by flipping a flag. # 2. The MoE Paradox (Bigger is Faster) How does a 26B parameter model absolutely destroy a 9B model in raw speed on the exact same hardware? Architecture. Qwen 3.5 9B is a dense model. it activates all 9 billion parameters for every single token. Gemma 4 26B is a Mixture of Experts (MoE) model. It routes data efficiently, activating only 4B parameters per token. You get the reasoning capabilities of a 26B model with the compute cost of a 4B model. 3. Thinking Efficiency When I ran the exact same complex prompt on both models, the larger MoE spent significantly fewer "thinking" tokens to arrive at the correct answer. A smarter model doesn't just give better answers; it gets to the point faster, saving you compute cycles and preserving your context window. # Want to run this yourself? Here are the exact llama.cpp CLI commands. For Qwen (MTP is baked into the main model): ./llama-cli -m Qwen3.5-9B-UD-Q4_K_XL.gguf -p "Explain quantum computing." -n 2000 -c 8000 -ngl 99 -fa on --spec-type draft-mtp --spec-draft-n-max 4 --spec-draft-p-min 0.7 For Gemma (Using a separate lightweight draft model): ./llama-cli -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --model-draft mtp-gemma-4-26B-A4B-it.gguf -p "Explain quantum computing." -n 2000 -c 8000 -ngl 99 -fa on --spec-type draft-mtp --spec-draft-n-max 4 --spec-draft-p-min 0.7 Stop waiting for a $3,000 rig. Boot up Colab, pull these models, and start building your stack. I’ve put together a completely free, cell by cell Google Colab notebook that automates this entire workflow so you can test it yourself in 5 minutes and learn. Link to the notebook is in the comments below. Experiemt with different MTP parameters, context windows and post your results in the comments.

Alok

170,442 просмотров • 2 месяцев назад

This is the most hilarious thing I saw and did today Ran gemma-4-12B-coder-fable5-composer2.5-v1-GGUF locally with 8 GB VRAM at 20+ tok/sec Anthropic's Claude Fable 5 launched June 9. By June 12 it was banned. I can't access it. You can't either. But here's the twist: I'm running a model trained on its chain of thought at 20 tok/s on my RTX 4060 8GB. Locally. Offline. No cloud. No export control. Enter: Gemma4-12B-Coder GGUF (Q4_K_M) Base: Google's gemma-4-12B-it Fine-tuned on verifiable Python CoT data: - Primary: Composer 2.5 real reasoning traces (only passing solutions kept) - Auxiliary: Fable 5 used to redo the hard cases Composer missed. Every training example's reasoning led to code that actually ran. No hallucinated logic. Llama.cpp flags: -m gemma4-coding-Q4_K_M.gguf -cnv -ngl 44 -c 64000 -v (huggingface model link in comments) Flag breakdown: -ngl 44 → offload 44 layers to GPU (tune this for your VRAM) -c 64000 → 64K context window -cnv → conversation/chat mode -v → verbose output The irony writes itself. Anthropic spent weeks telling the world Fable 5 (mythos) is too powerful to release. Then released it. Then got banned from serving it, including their own researchers. Meanwhile: a Gemma 4 12B fine tune, trained on Fable 5's reasoning, runs fully offline on my mid range consumer GPU No API. No cloud. Just me and llama.cpp. This is why local AI matters. Check out the model's link in the comments. How's your experience been with this model?
0:15

This is the most hilarious thing I saw and did today Ran gemma-4-12B-coder-fable5-composer2.5-v1-GGUF locally with 8 GB VRAM at 20+ tok/sec Anthropic's Claude Fable 5 launched June 9. By June 12 it was banned. I can't access it. You can't either. But here's the twist: I'm running a model trained on its chain of thought at 20 tok/s on my RTX 4060 8GB. Locally. Offline. No cloud. No export control. Enter: Gemma4-12B-Coder GGUF (Q4_K_M) Base: Google's gemma-4-12B-it Fine-tuned on verifiable Python CoT data: - Primary: Composer 2.5 real reasoning traces (only passing solutions kept) - Auxiliary: Fable 5 used to redo the hard cases Composer missed. Every training example's reasoning led to code that actually ran. No hallucinated logic. Llama.cpp flags: -m gemma4-coding-Q4_K_M.gguf -cnv -ngl 44 -c 64000 -v (huggingface model link in comments) Flag breakdown: -ngl 44 → offload 44 layers to GPU (tune this for your VRAM) -c 64000 → 64K context window -cnv → conversation/chat mode -v → verbose output The irony writes itself. Anthropic spent weeks telling the world Fable 5 (mythos) is too powerful to release. Then released it. Then got banned from serving it, including their own researchers. Meanwhile: a Gemma 4 12B fine tune, trained on Fable 5's reasoning, runs fully offline on my mid range consumer GPU No API. No cloud. Just me and llama.cpp. This is why local AI matters. Check out the model's link in the comments. How's your experience been with this model?

Alok

575,300 просмотров • 3 месяцев назад

you're paying $20/mo for something your $500 GPU can already do. Gemma 4 26B A4B QAT MoE + Hermes Agent running on a single RTX 4060 (8GB VRAM). Built a vision capable, 100% free, 100% local, private AI assistant that lives in my Chrome browser. No API keys. No cloud. No subscriptions. 100% vibe coded. 0% handholding. It has full context of whatever's on my screen can answer questions, summarize pages, extract data, and see images. Same local model handles everything, no external calls, ever. keep reading for the model and hermes agent tips i learnt while building this locally. Here's the exact setup for anyone running local LLMs on 6-8 GB VRAM: llama.cpp server flags (on my NVIDIA RTX 4060 8gb VRAM): -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --cache-type-k q8_0 --cache-type-v q8_0 -c 150000 --port 8080 Throughput with quantization: Prefill: 200-250 tokens/sec Decode: 20-25 tokens/sec reduce context if oom on 6 gb vram card. Key learnings: - Quantize KV cache to q8 for faster prefill/decode. Prefill goes from 100-150 (unquantized) to 200-250 tok/s (q8). - But watch out, once actual context grows past ~50k tokens on high entropy workloads, q8 KV quantization can cause hallucinations. Low entropy workloads are mostly unaffected. If you see it happening, drop the quantization. This is common across all local models. - In Hermes Agent settings -> Memory & Context, bump compression threshold from default 0.5 to 0.7. Default triggers way too frequent context compression and eats time. Up next: add persistent memory, web search, tool calling, streaming output and whatever you suggest. Running a 26B MoE with vision + 150k context window on 8GB VRAM would've sounded impossible 6 months ago. Works the same on the NVIDIA RTX 3060 Ti, 3070, 4060 Ti, 5060, 2080, or any 8GB card. VRAM is the only requirement. Local AI agents are closer than people think. You just need to know where the knobs are. Model's Unsloth quant hugging face link in the comments. Have you tried Hermes agent by Nous Research yet? What are you building with local LLMs? Drop it below, let's see what this community is shipping.
0:19

you're paying $20/mo for something your $500 GPU can already do. Gemma 4 26B A4B QAT MoE + Hermes Agent running on a single RTX 4060 (8GB VRAM). Built a vision capable, 100% free, 100% local, private AI assistant that lives in my Chrome browser. No API keys. No cloud. No subscriptions. 100% vibe coded. 0% handholding. It has full context of whatever's on my screen can answer questions, summarize pages, extract data, and see images. Same local model handles everything, no external calls, ever. keep reading for the model and hermes agent tips i learnt while building this locally. Here's the exact setup for anyone running local LLMs on 6-8 GB VRAM: llama.cpp server flags (on my NVIDIA RTX 4060 8gb VRAM): -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --cache-type-k q8_0 --cache-type-v q8_0 -c 150000 --port 8080 Throughput with quantization: Prefill: 200-250 tokens/sec Decode: 20-25 tokens/sec reduce context if oom on 6 gb vram card. Key learnings: - Quantize KV cache to q8 for faster prefill/decode. Prefill goes from 100-150 (unquantized) to 200-250 tok/s (q8). - But watch out, once actual context grows past ~50k tokens on high entropy workloads, q8 KV quantization can cause hallucinations. Low entropy workloads are mostly unaffected. If you see it happening, drop the quantization. This is common across all local models. - In Hermes Agent settings -> Memory & Context, bump compression threshold from default 0.5 to 0.7. Default triggers way too frequent context compression and eats time. Up next: add persistent memory, web search, tool calling, streaming output and whatever you suggest. Running a 26B MoE with vision + 150k context window on 8GB VRAM would've sounded impossible 6 months ago. Works the same on the NVIDIA RTX 3060 Ti, 3070, 4060 Ti, 5060, 2080, or any 8GB card. VRAM is the only requirement. Local AI agents are closer than people think. You just need to know where the knobs are. Model's Unsloth quant hugging face link in the comments. Have you tried Hermes agent by Nous Research yet? What are you building with local LLMs? Drop it below, let's see what this community is shipping.

Alok

36,691 просмотров • 2 месяцев назад

If you have an RTX 3090 or 4090, Mia just shipped you a free massive upgrade in both speed and intelligence. I will explain to you why this will make your Qwen 3.8 27B on your card, even better, and my flags for running it. Qwen3.8-27B, EXL3 3.5bpw, DFlash2 speculative decode, RTX 4090. Single stream. The kit is from MiaAI-Lab, EXL3 is turboderp's format. I re-measured everything on my own card because the my first benchmarks seemed off. It turns out it really does run much faster. WHY EXL3 IS A DIFFERENT ANIMAL The old way (Q4_K_M) rounds each weight to the nearest 4-bit value independently. Every weight introduces its own rounding error. Those errors accumulate across millions of weights and causes drift (Slightly dumber). EXL3 is a fundamentally different compression algorithm. Instead of rounding each weight on its own, it encodes the entire weight vector as a path through a constrained codebook and spreads the rounding error across dimensions using a Hadamard transform. The result is that at the same bits per weight, more of the original model's intelligence is preserved. The important part is this CAN ACTUALLY BE MEASURED. The cleanest way to see that is KL divergence against a high-precision teacher. Lower means the quantized model thinks more like the original. On the malaiwah independent teacher-logit panel for GLM-5.3-Flash: EXL3 4bpw: 0.0246 nats Official FP8: 0.0206 nats NVFP4: 0.0605 nats EXL3 sits 0.004 nats behind native FP8 at half the size. NVFP4 at higher bit width is 2.5x further from the teacher. That panel is GLM-5.3-Flash, not Qwen 3.8. Cited as the mechanism, not as this run's data. But the point stands: EXL3 is not just smaller, it is smarter per bit than the formats most people are running. WHAT I MEASURED I first measure 108 tok/s from a single run. After that number looked too good to be true. I reran it. It looks like after a warm up, the numbers are even better. Basically, like people long thought, the RTX 3090 and RTX 4090 are actually superb AI computer cards. Hence why NVIDIA stopped shipping them with NVLINK since the 4090. Short context ceiling (~2k in, 1016-token output, TTFT-separated): 135, 138, 153, 174, 133, 129 tok/s across 6 runs. Sustained longform (2040-token essay): 105.4, 94.5, 98.4 tok/s Short answer (504 tokens): 93.2 tok/s The honest shape: ~130-150 tok/s at short context is the ceiling, ~94-105 sustained on longform. The ceiling matters because that is what people feel in chat. The old dense Q4_K_M on llama.cpp ran ~37 tok/s on this same card. (No MTP), with MTP about 60 tok/s Sustained is roughly 2.5-3x. Ceiling is closer to 4x. Same model, different quantization and engine. The multiplier comes from EXL3, the ExLlamaV2 engine, and DFlash2 together. CONCURRENCY IS A RTX 4090 LANE. Just like the old config on the 4090, the 24gb vram, means a long context can only hold one stream, and running concurrency requires to lower context length, because it runs fast it sort of makes up for it by being faster than slower GPU chips. CONTEXT LADDER The recipe doc measured prefill. I re-ran it with TTFT separated from decode, because decode is what you actually feel after the first token. ~5k in: decode 140 tok/s (TTFT 0.5s), needle HIT ~18k in: decode 85 tok/s (TTFT 0.3s*), needle HIT ~73k in: decode 28 tok/s (TTFT 1.8s), needle HIT ~146k in: decode 16 tok/s (TTFT 2.3s), needle HIT (*0.3s at 18k is a prefix-cache hit from the paired pass. Cold prefill for reference: ~2,020 tok/s at 17k falling to ~508 at 153k.) Needle hit at every depth, mine and the original 7/7. Retrieval is intact at max context. Speed is not: decode falls ~9x from short to max. Past ~50k tokens this stops being a chat tool and becomes a batch tool. At 146k it works, but nobody is typing interactively against 16 tok/s. WHERE IT BROKE The model's native context is 262k. The README says DFlash2 fits ~220k on a 24GB card. My 200,704-token attempt failed with insufficient VRAM. Dropped to 168,960 and it booted. The real ceiling is somewhere between 168,960 and 200,704 and I never tested that gap. I jumped to a value that worked and called it done. That is ~32k tokens of context I left on the table. One thing the numbers taught me: JSON tokenizes at ~1.5 chars/token, prose at ~3.9. "150k tokens of JSON" needs ~2.7x more filler than the same estimate in prose. Size by real tokens, not estimates. THE UPGRADE If you own a 4090 and you are running Q4_K_M on llama.cpp, you are leaving a good bit of speed and measurable intelligence on the table. The same model, on the same card, with a better quantization and engine, goes from 60 tok/s to 130-150 at short context and 94-105 sustained. The model also thinks closer to the original because EXL3 preserves more of the output distribution per bit than the old rounding method. The recipe is in the first reply. Everything above came from one 4090 and one afternoon of re-measuring. The decode ladder especially needs independent numbers. If your card gets different falloff, that is worth knowing. Recipe and flags/ findings in reply 👇
0:15

If you have an RTX 3090 or 4090, Mia just shipped you a free massive upgrade in both speed and intelligence. I will explain to you why this will make your Qwen 3.8 27B on your card, even better, and my flags for running it. Qwen3.8-27B, EXL3 3.5bpw, DFlash2 speculative decode, RTX 4090. Single stream. The kit is from MiaAI-Lab, EXL3 is turboderp's format. I re-measured everything on my own card because the my first benchmarks seemed off. It turns out it really does run much faster. WHY EXL3 IS A DIFFERENT ANIMAL The old way (Q4_K_M) rounds each weight to the nearest 4-bit value independently. Every weight introduces its own rounding error. Those errors accumulate across millions of weights and causes drift (Slightly dumber). EXL3 is a fundamentally different compression algorithm. Instead of rounding each weight on its own, it encodes the entire weight vector as a path through a constrained codebook and spreads the rounding error across dimensions using a Hadamard transform. The result is that at the same bits per weight, more of the original model's intelligence is preserved. The important part is this CAN ACTUALLY BE MEASURED. The cleanest way to see that is KL divergence against a high-precision teacher. Lower means the quantized model thinks more like the original. On the malaiwah independent teacher-logit panel for GLM-5.3-Flash: EXL3 4bpw: 0.0246 nats Official FP8: 0.0206 nats NVFP4: 0.0605 nats EXL3 sits 0.004 nats behind native FP8 at half the size. NVFP4 at higher bit width is 2.5x further from the teacher. That panel is GLM-5.3-Flash, not Qwen 3.8. Cited as the mechanism, not as this run's data. But the point stands: EXL3 is not just smaller, it is smarter per bit than the formats most people are running. WHAT I MEASURED I first measure 108 tok/s from a single run. After that number looked too good to be true. I reran it. It looks like after a warm up, the numbers are even better. Basically, like people long thought, the RTX 3090 and RTX 4090 are actually superb AI computer cards. Hence why NVIDIA stopped shipping them with NVLINK since the 4090. Short context ceiling (~2k in, 1016-token output, TTFT-separated): 135, 138, 153, 174, 133, 129 tok/s across 6 runs. Sustained longform (2040-token essay): 105.4, 94.5, 98.4 tok/s Short answer (504 tokens): 93.2 tok/s The honest shape: ~130-150 tok/s at short context is the ceiling, ~94-105 sustained on longform. The ceiling matters because that is what people feel in chat. The old dense Q4_K_M on llama.cpp ran ~37 tok/s on this same card. (No MTP), with MTP about 60 tok/s Sustained is roughly 2.5-3x. Ceiling is closer to 4x. Same model, different quantization and engine. The multiplier comes from EXL3, the ExLlamaV2 engine, and DFlash2 together. CONCURRENCY IS A RTX 4090 LANE. Just like the old config on the 4090, the 24gb vram, means a long context can only hold one stream, and running concurrency requires to lower context length, because it runs fast it sort of makes up for it by being faster than slower GPU chips. CONTEXT LADDER The recipe doc measured prefill. I re-ran it with TTFT separated from decode, because decode is what you actually feel after the first token. ~5k in: decode 140 tok/s (TTFT 0.5s), needle HIT ~18k in: decode 85 tok/s (TTFT 0.3s*), needle HIT ~73k in: decode 28 tok/s (TTFT 1.8s), needle HIT ~146k in: decode 16 tok/s (TTFT 2.3s), needle HIT (*0.3s at 18k is a prefix-cache hit from the paired pass. Cold prefill for reference: ~2,020 tok/s at 17k falling to ~508 at 153k.) Needle hit at every depth, mine and the original 7/7. Retrieval is intact at max context. Speed is not: decode falls ~9x from short to max. Past ~50k tokens this stops being a chat tool and becomes a batch tool. At 146k it works, but nobody is typing interactively against 16 tok/s. WHERE IT BROKE The model's native context is 262k. The README says DFlash2 fits ~220k on a 24GB card. My 200,704-token attempt failed with insufficient VRAM. Dropped to 168,960 and it booted. The real ceiling is somewhere between 168,960 and 200,704 and I never tested that gap. I jumped to a value that worked and called it done. That is ~32k tokens of context I left on the table. One thing the numbers taught me: JSON tokenizes at ~1.5 chars/token, prose at ~3.9. "150k tokens of JSON" needs ~2.7x more filler than the same estimate in prose. Size by real tokens, not estimates. THE UPGRADE If you own a 4090 and you are running Q4_K_M on llama.cpp, you are leaving a good bit of speed and measurable intelligence on the table. The same model, on the same card, with a better quantization and engine, goes from 60 tok/s to 130-150 at short context and 94-105 sustained. The model also thinks closer to the original because EXL3 preserves more of the output distribution per bit than the old rounding method. The recipe is in the first reply. Everything above came from one 4090 and one afternoon of re-measuring. The decode ladder especially needs independent numbers. If your card gets different falloff, that is worth knowing. Recipe and flags/ findings in reply 👇

Yume_X

38,511 просмотров • 10 дней назад

Testing the new Gemma 4 12B (QAT) vision and OCR capabilities locally with LM Studio. # The setup: - GPU: NVIDIA RTX 4060 (8GB VRAM) - CPU: Intel i7 - Runner: LM Studio - Config: 32k context, 38 layers offloaded, Flash Attention enabled - Speed: ~14 tokens/sec decode throughput # The test: I gave it a screenshot of Google AI Studio. Prompt: "clone this. give me a single html file" # The result: A solid one shot replication. It successfully mapped out the layout, recognized the UI text, and structured the divs correctly, with only minor differences from the original. Results available at the end of the video. Quite capable for a 12B model running on budget consumer hardware. A gpu that costs only $300. # Why the architecture under the hood is notable: Unlike traditional models that rely on heavy, separate vision and audio encoders, Gemma 4 12B uses a unified, encoder free architecture. It bypasses separate multi stage encoders. Uses a 35M parameter vision embedder to project raw 48x48 pixel patches directly to the LLM hidden dimension. Local multimodal development is becoming highly accessible on standard hardware. If you've spun up Gemma 4 12B locally, what setup are you using and what kind of throughput are you seeing?
0:16

Testing the new Gemma 4 12B (QAT) vision and OCR capabilities locally with LM Studio. # The setup: - GPU: NVIDIA RTX 4060 (8GB VRAM) - CPU: Intel i7 - Runner: LM Studio - Config: 32k context, 38 layers offloaded, Flash Attention enabled - Speed: ~14 tokens/sec decode throughput # The test: I gave it a screenshot of Google AI Studio. Prompt: "clone this. give me a single html file" # The result: A solid one shot replication. It successfully mapped out the layout, recognized the UI text, and structured the divs correctly, with only minor differences from the original. Results available at the end of the video. Quite capable for a 12B model running on budget consumer hardware. A gpu that costs only $300. # Why the architecture under the hood is notable: Unlike traditional models that rely on heavy, separate vision and audio encoders, Gemma 4 12B uses a unified, encoder free architecture. It bypasses separate multi stage encoders. Uses a 35M parameter vision embedder to project raw 48x48 pixel patches directly to the LLM hidden dimension. Local multimodal development is becoming highly accessible on standard hardware. If you've spun up Gemma 4 12B locally, what setup are you using and what kind of throughput are you seeing?

Alok

25,717 просмотров • 3 месяцев назад

hey if you have a 3060, or any GPU with 8GB or more sitting in a drawer right now, that thing can run 9 billion parameters of intelligence autonomously. and you don't know it yet. 2 hours ago i posted that 9B hit a ceiling. 2,699 lines across 11 files. blank screen. said the limit for autonomous multifile coding on 9 billion parameters is real. then i audited every file. found 11 bugs. exact file, exact line, exact fix. duplicate variable declarations killing the script loader. a canvas reference never connected to the DOM. enemies with no movement logic. particle systems called on the class instead of the instance. fed that list as a single prompt to the same Qwen 3.5 9B on the same RTX 3060 through Hermes Agent. it fixed all 11. surgically. patch level edits across 4 files. no rewrites. no hallucinated changes. game boots. enemies spawn, move, collide. background renders. particles fire. and here's what nobody is talking about. this is a 9 billion parameter model running a full agentic framework. Hermes Agent with 31 tools. file operations, terminal, browser, code execution. not a single tool call failed. the agent chain never broke. most people think you need 70B+ for reliable tool use. this is 9B on 12 gigs doing it clean. the model didn't fail. my prompting strategy did. the ceiling is not the parameter count. the ceiling is how you prompt it. this is not done. bullets don't fire yet. boss fights need wiring. but the screen that was black 2 hours ago now has a full game rendering in real time. iterating right now. anyone with a GPU from the last 5 years should be paying attention to what is happening right now.
0:12

hey if you have a 3060, or any GPU with 8GB or more sitting in a drawer right now, that thing can run 9 billion parameters of intelligence autonomously. and you don't know it yet. 2 hours ago i posted that 9B hit a ceiling. 2,699 lines across 11 files. blank screen. said the limit for autonomous multifile coding on 9 billion parameters is real. then i audited every file. found 11 bugs. exact file, exact line, exact fix. duplicate variable declarations killing the script loader. a canvas reference never connected to the DOM. enemies with no movement logic. particle systems called on the class instead of the instance. fed that list as a single prompt to the same Qwen 3.5 9B on the same RTX 3060 through Hermes Agent. it fixed all 11. surgically. patch level edits across 4 files. no rewrites. no hallucinated changes. game boots. enemies spawn, move, collide. background renders. particles fire. and here's what nobody is talking about. this is a 9 billion parameter model running a full agentic framework. Hermes Agent with 31 tools. file operations, terminal, browser, code execution. not a single tool call failed. the agent chain never broke. most people think you need 70B+ for reliable tool use. this is 9B on 12 gigs doing it clean. the model didn't fail. my prompting strategy did. the ceiling is not the parameter count. the ceiling is how you prompt it. this is not done. bullets don't fire yet. boss fights need wiring. but the screen that was black 2 hours ago now has a full game rendering in real time. iterating right now. anyone with a GPU from the last 5 years should be paying attention to what is happening right now.

Sudo su

684,336 просмотров • 6 месяцев назад

i've been paying 200$/month for AI tools for a while now claude, chatgpt, all of it thought that was just the price of doing things properly then i read this and had to close my laptop for a second someone bought a used GPU from eBay for $700 five years old. RTX 3090. the kind of card people sell when they upgrade for gaming plugged in a free model called Qwen 3.6 that alibaba quietly dropped last month and it scored 84.1 on vision tasks claude 4.5 opus scored 77.0 the 200$/month subscription lost to a free model running on old hardware the math gets worse after 4 months you've paid off the GPU from month 5 onwards you pay $8 for electricity that's it no subscription. no per message cost. nothing leaves your computer. ever. turns out i didn't have to is anyone else just finding out about this or was i the last one to know?
0:15

i've been paying 200$/month for AI tools for a while now claude, chatgpt, all of it thought that was just the price of doing things properly then i read this and had to close my laptop for a second someone bought a used GPU from eBay for $700 five years old. RTX 3090. the kind of card people sell when they upgrade for gaming plugged in a free model called Qwen 3.6 that alibaba quietly dropped last month and it scored 84.1 on vision tasks claude 4.5 opus scored 77.0 the 200$/month subscription lost to a free model running on old hardware the math gets worse after 4 months you've paid off the GPU from month 5 onwards you pay $8 for electricity that's it no subscription. no per message cost. nothing leaves your computer. ever. turns out i didn't have to is anyone else just finding out about this or was i the last one to know?

Crypto Mavka

14,126 просмотров • 2 месяцев назад

Open source AI is actually moving at an unhinged pace right now. I literally hadn't even finished typing up my last Gemma 4 12b benchmark notes before Google went ahead and dropped the official Quantization Aware Training (QAT) checkpoints on Hugging Face. If you missed the news, QAT basically bakes the compression directly into the training process. Instead of standard post training quantization degrading the model's reasoning capabilities, QAT trains the model with compression in mind. Unsloth is reporting near original performance at 4-bit with ~72% lower memory footprint. Details in the comments. Naturally, had to instantly pull the new GGUFs to see what a single RTX 4090 card (24 GB VRAM, Cuda 12.8, ubuntu 22) could do. i fired up llama.cpp engine again Look at these numbers: 1. Unsloth Gemma 4 26B-A4B IT (QAT Q4_K_XL) flags: ./build/bin/llama-cli -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf -cnv -ngl 99 -c 250000 -fa on -v VRAM Used: 19.5 GB context: 250,000 tokens decode throughput: 193 tps 2. Unsloth Gemma 4 31B IT (QAT Q4_K_XL) flags: Command: ./build/bin/llama-cli -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -cnv -ngl 99 -c 60000 -fa on -v - VRAM Used: 23 GB (Tight, but zero system RAM spillover) - context: 60,000 tokens - decode throughput: 47 tps We are essentially watching hardware bottlenecks evaporate in real time. An update literally drops before you can finish benchmarking the previous one. What a time to be running local hardware. If you have a single rtx 3090, rtx 4090, these are the latest gemma models to try this week.
0:14

Open source AI is actually moving at an unhinged pace right now. I literally hadn't even finished typing up my last Gemma 4 12b benchmark notes before Google went ahead and dropped the official Quantization Aware Training (QAT) checkpoints on Hugging Face. If you missed the news, QAT basically bakes the compression directly into the training process. Instead of standard post training quantization degrading the model's reasoning capabilities, QAT trains the model with compression in mind. Unsloth is reporting near original performance at 4-bit with ~72% lower memory footprint. Details in the comments. Naturally, had to instantly pull the new GGUFs to see what a single RTX 4090 card (24 GB VRAM, Cuda 12.8, ubuntu 22) could do. i fired up llama.cpp engine again Look at these numbers: 1. Unsloth Gemma 4 26B-A4B IT (QAT Q4_K_XL) flags: ./build/bin/llama-cli -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf -cnv -ngl 99 -c 250000 -fa on -v VRAM Used: 19.5 GB context: 250,000 tokens decode throughput: 193 tps 2. Unsloth Gemma 4 31B IT (QAT Q4_K_XL) flags: Command: ./build/bin/llama-cli -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -cnv -ngl 99 -c 60000 -fa on -v - VRAM Used: 23 GB (Tight, but zero system RAM spillover) - context: 60,000 tokens - decode throughput: 47 tps We are essentially watching hardware bottlenecks evaporate in real time. An update literally drops before you can finish benchmarking the previous one. What a time to be running local hardware. If you have a single rtx 3090, rtx 4090, these are the latest gemma models to try this week.

Alok

26,841 просмотров • 3 месяцев назад

Fine-tune DeepSeek-OCR on your own language! (100% local) DeepSeek-OCR is a 3B-parameter vision model that achieves 97% precision while using 10× fewer vision tokens than text-based LLMs. It handles tables, papers, and handwriting without killing your GPU or budget. Why it matters: Most vision models treat documents as massive sequences of tokens, making long-context processing expensive and slow. DeepSeek-OCR uses context optical compression to convert 2D layouts into vision tokens, enabling efficient processing of complex documents. The best part? You can easily fine-tune it for your specific use case on a single GPU. I used Unsloth to run this experiment on Persian text and saw an 88.26% improvement in character error rate. ↳ Base model: 149% character error rate (CER) ↳ Fine-tuned model: 60% CER (57% more accurate) ↳ Training time: 60 steps on a single GPU Persian was just the test case. You can swap in your own dataset for any language, document type, or specific domain you're working with. I've shared the complete guide in the next tweet - all the code, notebooks, and environment setup ready to run with a single click. Everything is 100% open-source!
0:14

Fine-tune DeepSeek-OCR on your own language! (100% local) DeepSeek-OCR is a 3B-parameter vision model that achieves 97% precision while using 10× fewer vision tokens than text-based LLMs. It handles tables, papers, and handwriting without killing your GPU or budget. Why it matters: Most vision models treat documents as massive sequences of tokens, making long-context processing expensive and slow. DeepSeek-OCR uses context optical compression to convert 2D layouts into vision tokens, enabling efficient processing of complex documents. The best part? You can easily fine-tune it for your specific use case on a single GPU. I used Unsloth to run this experiment on Persian text and saw an 88.26% improvement in character error rate. ↳ Base model: 149% character error rate (CER) ↳ Fine-tuned model: 60% CER (57% more accurate) ↳ Training time: 60 steps on a single GPU Persian was just the test case. You can swap in your own dataset for any language, document type, or specific domain you're working with. I've shared the complete guide in the next tweet - all the code, notebooks, and environment setup ready to run with a single click. Everything is 100% open-source!

Akshay 🚀

126,213 просмотров • 10 месяцев назад

90% of "AI developers" just download pre packaged GGUF files from Hugging Face, hit run, and call it a day. The top 10% know how to pull the raw safetensors, run the math, and quantize massive models into Q4_K_M themselves. If you think llama.cpp can only execute models, you’re missing the best part of the open source ecosystem. It’s a high performance optimization suite. Manually stripping 69% of the VRAM footprint off a brand new model architecture is where real infrastructure value is made. If you want to actually master local inference and deploy models like Google’s massive Gemma 4 12B it on consumer NVIDIA hardware using llama.cpp, you need to learn this pipeline. Let's build it. I just took the raw 22.7 GB Gemma 4 baseline and manually compressed it down to a 7.02 GB Q4_K_M GGUF artifact using llama.cpp. That is a 69% reduction in footprint. No quality loss. No VRAM bottlenecks. Just native, hardware accelerated C++ inference running a full 2,50,000 token context window on a dual NVIDIA Tesla T4 setup. Stop melting your VRAM on unoptimized weights and stop relying on other people's pipelines. Own your stack. I mapped this entire architecture from dynamic binary fetching to raw quantization and real time GPU streaming into a single, bulletproof notebook. Notebook link is in the comments below. Bookmark this blueprint for your next deployment and tell me which quantization works best for your workflow and model.
0:19

90% of "AI developers" just download pre packaged GGUF files from Hugging Face, hit run, and call it a day. The top 10% know how to pull the raw safetensors, run the math, and quantize massive models into Q4_K_M themselves. If you think llama.cpp can only execute models, you’re missing the best part of the open source ecosystem. It’s a high performance optimization suite. Manually stripping 69% of the VRAM footprint off a brand new model architecture is where real infrastructure value is made. If you want to actually master local inference and deploy models like Google’s massive Gemma 4 12B it on consumer NVIDIA hardware using llama.cpp, you need to learn this pipeline. Let's build it. I just took the raw 22.7 GB Gemma 4 baseline and manually compressed it down to a 7.02 GB Q4_K_M GGUF artifact using llama.cpp. That is a 69% reduction in footprint. No quality loss. No VRAM bottlenecks. Just native, hardware accelerated C++ inference running a full 2,50,000 token context window on a dual NVIDIA Tesla T4 setup. Stop melting your VRAM on unoptimized weights and stop relying on other people's pipelines. Own your stack. I mapped this entire architecture from dynamic binary fetching to raw quantization and real time GPU streaming into a single, bulletproof notebook. Notebook link is in the comments below. Bookmark this blueprint for your next deployment and tell me which quantization works best for your workflow and model.

Alok

62,631 просмотров • 2 месяцев назад

this is the worst local ai will ever be. it only gets better from here. if you are not expanding your mind with these small models you are missing what's happening right now 99 percent tool call success rate. when steered well with the right skills and a framework like hermes agent the node becomes a cognition layer. not a chatbot. not a toy. an extension of how you think. i was cranking this node at 35 to 50 tok/s all day on personal experiments and now after all the work is done qwen 3.5 9B is iterating on its own code. the game it created. fixing its own bugs autonomously. and the part you should probably not miss is that all of this is happening on a RTX 3060. not an H100. not an A100. the card most of you have sitting in a drawer right now. if you just open that drawer and put that intelligence to work every tensor core on that card should be running for you. your work. your experiments. your thinking. you all have it but because nobody told you what this hardware can actually do in 2026 you never tried. the day it unlocks is the day you test your workload, understand the tradeoffs, debug the loops, and then decide if you need to scale the hardware. there is no point buying 3 mac studios when things done well you can squeeze a similar level of intelligence from 9B compared to 70B. but only when you create the right environment for your model through the right harness. and let me tell you i have tried claude code as a local harness. i have tried opencode. i have tried various others. somehow i landed on hermes agent and never left. there is something magical going on at Nous Research. the tool call parsers, the skills system, the way it handles small models natively. nothing else comes close for local inference. own your cognition. your AI. your agent. your prompts. your experiments. why give them away for free. those are who you are and they don't belong on someone else's servers being monitored. just give it a shot with your existing hardware. you run into a problem the community will help you. and if you are migrating from openclaw to hermes i will personally help you make the switch.
0:13

this is the worst local ai will ever be. it only gets better from here. if you are not expanding your mind with these small models you are missing what's happening right now 99 percent tool call success rate. when steered well with the right skills and a framework like hermes agent the node becomes a cognition layer. not a chatbot. not a toy. an extension of how you think. i was cranking this node at 35 to 50 tok/s all day on personal experiments and now after all the work is done qwen 3.5 9B is iterating on its own code. the game it created. fixing its own bugs autonomously. and the part you should probably not miss is that all of this is happening on a RTX 3060. not an H100. not an A100. the card most of you have sitting in a drawer right now. if you just open that drawer and put that intelligence to work every tensor core on that card should be running for you. your work. your experiments. your thinking. you all have it but because nobody told you what this hardware can actually do in 2026 you never tried. the day it unlocks is the day you test your workload, understand the tradeoffs, debug the loops, and then decide if you need to scale the hardware. there is no point buying 3 mac studios when things done well you can squeeze a similar level of intelligence from 9B compared to 70B. but only when you create the right environment for your model through the right harness. and let me tell you i have tried claude code as a local harness. i have tried opencode. i have tried various others. somehow i landed on hermes agent and never left. there is something magical going on at Nous Research. the tool call parsers, the skills system, the way it handles small models natively. nothing else comes close for local inference. own your cognition. your AI. your agent. your prompts. your experiments. why give them away for free. those are who you are and they don't belong on someone else's servers being monitored. just give it a shot with your existing hardware. you run into a problem the community will help you. and if you are migrating from openclaw to hermes i will personally help you make the switch.

Sudo su

58,717 просмотров • 6 месяцев назад

Muse Glimmer, A 30B parameter dense model swallowing a 130,000 token context window using only 19.3 GB of VRAM (extreme efficiency). No KV cache quantization required. I just benched the new Muse Glimmer 30B (dense) on a single RTX 4090. We are pulling 3,100+ t/s prefill and 75 tokens/second decode. The throughput is violent. Meta superintelligence lab just open sourced this agentic beast, explicitly engineered to dominate 24GB consumer cards. I pulled the latest llama.cpp source on Ubuntu 22 (CUDA 13) to see if the specs were real. Fed it a 28k token prompt. Here is the exact llama.cpp God Stack and benchmarking breakdown: # 1. The Deep Context Run (No Speculative Decoding) The architecture uses a massive 16:1 GQA (Grouped Query Attention) ratio. This means the KV cache footprint is practically non existent. ./build/bin/llama-server -m Muse-Glimmer-30B-UD-Q4_K_XL.gguf -c 130000 -b 4096 -ub 4096 -ngl 99 --port 8080 Prefill: 3134.95 t/s Decode: 50.00 t/s VRAM: 19.34 GB (I hit 130k context on pristine, unquantized f16 cache and still had 4.5 GB of VRAM left over. Absolute witchcraft). # 2. The DFlash Speculative Overdrive Meta shipped this with a DFlash block diffusion drafter. Let's trade that extra VRAM for pure speed. ./build/bin/llama-server -m Muse-Glimmer-30B-UD-Q4_K_XL.gguf -md dflash-kquant.gguf --spec-type draft-dflash --spec-draft-n-max 3 -c 80000 -b 4096 -ub 4096 -ngl 99 --port 8080 Prefill: 1293.69 t/s Decode: 75.00 t/s VRAM: 23.93 GB (Maxed out on card) the dflash gguf is additional 1.6 GBs # The Architecture Insight (Muse Glimmer vs. Gemma 4 31B) If you look at my Gemma 4 31B tests from last week, getting 140k context required heavily degrading the memory with Q4 KV quantization (gemma 31b q4 can do only about 40k context with unquantized kv on a 24gb card). That "unzipping" overhead bottlenecked Gemma's MTP decode speeds down to 65 t/s. Muse Glimmer completely sidesteps this bottleneck. By using aggressive 16:1 GQA, it keeps the KV cache in native f16 format at massive context lengths. Flash Attention gets to run at maximum uncompressed speed, letting the DFlash drafter push decode safely to 75 t/s without compute lag. With a 76% on SWE Bench Verified and seamless local tool calling, this model looks promising. Unsloth's Hugging Face GGUF links, intelligence/agentic benchmark details, and inference throughput performance graphs are posted in the replies. For 24GB rig, what’s your current go to model?
0:16

Muse Glimmer, A 30B parameter dense model swallowing a 130,000 token context window using only 19.3 GB of VRAM (extreme efficiency). No KV cache quantization required. I just benched the new Muse Glimmer 30B (dense) on a single RTX 4090. We are pulling 3,100+ t/s prefill and 75 tokens/second decode. The throughput is violent. Meta superintelligence lab just open sourced this agentic beast, explicitly engineered to dominate 24GB consumer cards. I pulled the latest llama.cpp source on Ubuntu 22 (CUDA 13) to see if the specs were real. Fed it a 28k token prompt. Here is the exact llama.cpp God Stack and benchmarking breakdown: # 1. The Deep Context Run (No Speculative Decoding) The architecture uses a massive 16:1 GQA (Grouped Query Attention) ratio. This means the KV cache footprint is practically non existent. ./build/bin/llama-server -m Muse-Glimmer-30B-UD-Q4_K_XL.gguf -c 130000 -b 4096 -ub 4096 -ngl 99 --port 8080 Prefill: 3134.95 t/s Decode: 50.00 t/s VRAM: 19.34 GB (I hit 130k context on pristine, unquantized f16 cache and still had 4.5 GB of VRAM left over. Absolute witchcraft). # 2. The DFlash Speculative Overdrive Meta shipped this with a DFlash block diffusion drafter. Let's trade that extra VRAM for pure speed. ./build/bin/llama-server -m Muse-Glimmer-30B-UD-Q4_K_XL.gguf -md dflash-kquant.gguf --spec-type draft-dflash --spec-draft-n-max 3 -c 80000 -b 4096 -ub 4096 -ngl 99 --port 8080 Prefill: 1293.69 t/s Decode: 75.00 t/s VRAM: 23.93 GB (Maxed out on card) the dflash gguf is additional 1.6 GBs # The Architecture Insight (Muse Glimmer vs. Gemma 4 31B) If you look at my Gemma 4 31B tests from last week, getting 140k context required heavily degrading the memory with Q4 KV quantization (gemma 31b q4 can do only about 40k context with unquantized kv on a 24gb card). That "unzipping" overhead bottlenecked Gemma's MTP decode speeds down to 65 t/s. Muse Glimmer completely sidesteps this bottleneck. By using aggressive 16:1 GQA, it keeps the KV cache in native f16 format at massive context lengths. Flash Attention gets to run at maximum uncompressed speed, letting the DFlash drafter push decode safely to 75 t/s without compute lag. With a 76% on SWE Bench Verified and seamless local tool calling, this model looks promising. Unsloth's Hugging Face GGUF links, intelligence/agentic benchmark details, and inference throughput performance graphs are posted in the replies. For 24GB rig, what’s your current go to model?

Alok

65,480 просмотров • 1 месяц назад

MiniMax H3 Instead of sharing the prompts for each of these videos, I thought it would be more useful to share how I created that prompts. All of the videos were generated with text-to-video. First, find an image with the kind of scene, composition and mood you want to recreate. I used a few YouTube playlist thumbnails as references but Pinterest is also a great place to find inspiration. You can even use your own old or nostalgic photographs. Then upload the image to ChatGPT and ask it to describe the scene. The description it gives you can essentially become your text-to-video prompt. From there, you can generate completely new scenes with a similar composition, atmosphere and cinematic language. You can of course use the reference image directly with image-to-video or as a first frame. But if the original image isn't yours, I prefer using it only as visual inspiration and recreating the scene through text-to-video. This is the prompt I use with ChatGPT: "Describe the scene in this image in English, focusing primarily on what is happening, the characters, their actions and body language, the setting and the overall atmosphere. Also briefly describe the composition, framing, camera angle, approximate lens choice, lighting, color palette and cinematic aesthetic. Keep it concise and scene-focused rather than overly technical."
0:15

MiniMax H3 Instead of sharing the prompts for each of these videos, I thought it would be more useful to share how I created that prompts. All of the videos were generated with text-to-video. First, find an image with the kind of scene, composition and mood you want to recreate. I used a few YouTube playlist thumbnails as references but Pinterest is also a great place to find inspiration. You can even use your own old or nostalgic photographs. Then upload the image to ChatGPT and ask it to describe the scene. The description it gives you can essentially become your text-to-video prompt. From there, you can generate completely new scenes with a similar composition, atmosphere and cinematic language. You can of course use the reference image directly with image-to-video or as a first frame. But if the original image isn't yours, I prefer using it only as visual inspiration and recreating the scene through text-to-video. This is the prompt I use with ChatGPT: "Describe the scene in this image in English, focusing primarily on what is happening, the characters, their actions and body language, the setting and the overall atmosphere. Also briefly describe the composition, framing, camera angle, approximate lens choice, lighting, color palette and cinematic aesthetic. Keep it concise and scene-focused rather than overly technical."

Kōda

55,398 просмотров • 29 дней назад

If you thought the Gemma 4 31B (dense) model was fast, sit down. I just benched the updated Gemma 4 26B A4B MoE on a single RTX 4090 (24 GB VRAM) 9,200 t/s prefill. 160 t/s decode. 250,000 context window. All on a single consumer RTX 4090. The numbers are completely unhinged. The 31B is a dense behemoth. But the 26B is a Mixture of Experts (MoE), specifically an Active 4 Billion (A4B). It holds 26B parameters of knowledge but only activates 4B per token. Because its inference memory footprint is so light, I didn’t even need KV cache quantization to hit a quarter million context. Compiled the latest llama.cpp from source on Ubuntu 22 (CUDA 13). Fed it a 28k token prompt, and manually cranked the batch sizes (-b 2048 -ub 2048) to absolutely redline the Tensor Cores. Here is the benchmarking breakdown: # 1. The Baseline (No MTP) Even without speculative decoding, the A4B architecture flies. llama.cpp flags: ./build/bin/llama-server -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf -c 250000 -ngl 99 -fa on -b 2048 -ub 2048 --port 8080 -v Context Ceiling: 250,000 tokens (21.5 GB VRAM) Prefill: 9,200 t/s (Absurd) Decode: 124 t/s # 2. The MTP Overdrive Injected the new MTP draft model to enable Speculative Decoding. llama.cpp flags: ./build/bin/llama-server -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --spec-type draft-mtp --spec-draft-model mtp-gemma-4-26B-A4B-it.gguf --spec-draft-n-max 4 --spec-draft-p-min 0.7 -c 250000 -ngl 99 -fa on -b 2048 -ub 2048 --port 8080 -v Context Ceiling: 250,000 tokens (22.96 GB VRAM) Prefill: 7,054 t/s (MTP draft overhead slightly caps prefill) Decode: 156 t/s # The Agentic Architecture Insight Why does this matter? Because you can now build a killer local agentic loop on a consumer desktop. Use the 31B dense model (from the previous post) as your heavy, deliberate Orchestrator / Verifier / Planner. Pass the actual execution tasks to this 26B MoE. At 160 t/s, this MoE can chew through code generation, tool calling, and massive RAG document retrieval over a 250k context window almost instantly, drastically speeding up your agentic loop. If you own a single RTX 3090 or 4090 and haven't tried this specific stack yet, you need to pull these latest updates and run it. Local inference just leveled up. Hugging Face links to the Unsloth 26B QAT quants and MTP drafters are in the replies. performance graphs also available in the replies.
0:17

If you thought the Gemma 4 31B (dense) model was fast, sit down. I just benched the updated Gemma 4 26B A4B MoE on a single RTX 4090 (24 GB VRAM) 9,200 t/s prefill. 160 t/s decode. 250,000 context window. All on a single consumer RTX 4090. The numbers are completely unhinged. The 31B is a dense behemoth. But the 26B is a Mixture of Experts (MoE), specifically an Active 4 Billion (A4B). It holds 26B parameters of knowledge but only activates 4B per token. Because its inference memory footprint is so light, I didn’t even need KV cache quantization to hit a quarter million context. Compiled the latest llama.cpp from source on Ubuntu 22 (CUDA 13). Fed it a 28k token prompt, and manually cranked the batch sizes (-b 2048 -ub 2048) to absolutely redline the Tensor Cores. Here is the benchmarking breakdown: # 1. The Baseline (No MTP) Even without speculative decoding, the A4B architecture flies. llama.cpp flags: ./build/bin/llama-server -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf -c 250000 -ngl 99 -fa on -b 2048 -ub 2048 --port 8080 -v Context Ceiling: 250,000 tokens (21.5 GB VRAM) Prefill: 9,200 t/s (Absurd) Decode: 124 t/s # 2. The MTP Overdrive Injected the new MTP draft model to enable Speculative Decoding. llama.cpp flags: ./build/bin/llama-server -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --spec-type draft-mtp --spec-draft-model mtp-gemma-4-26B-A4B-it.gguf --spec-draft-n-max 4 --spec-draft-p-min 0.7 -c 250000 -ngl 99 -fa on -b 2048 -ub 2048 --port 8080 -v Context Ceiling: 250,000 tokens (22.96 GB VRAM) Prefill: 7,054 t/s (MTP draft overhead slightly caps prefill) Decode: 156 t/s # The Agentic Architecture Insight Why does this matter? Because you can now build a killer local agentic loop on a consumer desktop. Use the 31B dense model (from the previous post) as your heavy, deliberate Orchestrator / Verifier / Planner. Pass the actual execution tasks to this 26B MoE. At 160 t/s, this MoE can chew through code generation, tool calling, and massive RAG document retrieval over a 250k context window almost instantly, drastically speeding up your agentic loop. If you own a single RTX 3090 or 4090 and haven't tried this specific stack yet, you need to pull these latest updates and run it. Local inference just leveled up. Hugging Face links to the Unsloth 26B QAT quants and MTP drafters are in the replies. performance graphs also available in the replies.

Alok

40,993 просмотров • 1 месяц назад