正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

🎉 Congrats to Thinking Machines on TML Inkling—a 1T-parameter open-weight model supported in vLLM from Day 0. Highlights: • Natively multimodal across text, image, and audio • Up to 1M-token context • New architecture with relative attention, short convolutions, and MoE expert sinks • 8 MTP heads for speculative... show more

vLLM

46,028 subscribers

49,079 次观看 • 23 天前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

🎉 Congrats to MiniMax (official) on releasing the open weights for MiniMax H3! Day-0 support in vLLM-Omni! One model reads text, images, video, and audio as a single context and returns video with native stereo audio. Text-to-video, first/last-frame, and multi-reference generation, 4 to 15 seconds at up to 2K and 24 FPS. The MP4 comes back with H.264 video and a synchronized stereo track already muxed in. It serves over the OpenAI-compatible /v1/videos endpoint, synchronous or async with job polling. 🔊 The video ⬇️ was made with H3, served on vLLM-Omni.

🎉 Congrats to MiniMax (official) on releasing the open weights for MiniMax H3! Day-0 support in vLLM-Omni! One model reads text, images, video, and audio as a single context and returns video with native stereo audio. Text-to-video, first/last-frame, and multi-reference generation, 4 to 15 seconds at up to 2K and 24 FPS. The MP4 comes back with H.264 video and a synchronized stereo track already muxed in. It serves over the OpenAI-compatible /v1/videos endpoint, synchronous or async with job polling. 🔊 The video ⬇️ was made with H3, served on vLLM-Omni.

vLLM

123,162 次观看 • 5 天前

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.

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 次观看 • 26 天前

Day 11/90 of Inference Engineering How does vLLM work and how is it used in production? Before we discuss how vLLM works internally, it helps to understand what vLLM is. At a high level, vLLM is an inference engine that is designed to serve LLMs to thousands of concurrent users efficiently while managing scarce compute and memory. The goal for vLLM is to maximize throughput and minimize latency; optimizing for the best inference economics and experience for end users. With every request from the end user, it eventually ends up in the engine core, gets scheduled alongside other requests from other concurrent users, executes on the GPU, and updates the KV cache with the new key and value vectors, and streams the tokens back to the user. The Scheduler decides what requests should execute next while continuously batching requests together to maximize GPU utilization. Continuous batching is an inference optimization that allows new requests to join a running batch as other requests finish generating tokens. This helps with keeping the GPU utilization high instead of letting it sit idle waiting for an entire batch to complete generating. After the scheduler dispatches the selected batch to the Model Executor, the Model Executor prepares the tensors and metadata required for inference, retrieves each request’s block table from KV Cache Manager, launches the optimized transformer forward pass on the GPU, computes the logits, updates the KV cache with the new key and value vectors, and finally returns the results for sampling and streaming. The KV Cache Manager uses the PagedAttention memory layout to allocate fixed-size cache blocks on demand and maintains a Free Block Queue on the CPU that tracks which blocks in the GPU’s Paged KV Cache are currently free. When a request needs additional KV cache space, the KV Cache manager takes a free block from the queue and assigns it to that request, thus avoiding an expensive search through GPU memory for available cache blocks. All of these components form the core of vLLM’s inference engine. The Scheduler determines what requests are executed, the Model Executor determines how those requests are executed, the KV Cache Manager determines where each request’s KV cache lives using the PagedAttention Memory Layout. This architecture enables vLLM to serve thousands of concurrent requests with high throughput, low latency, and efficient GPU memory utilization. Heres a little animation that visualizes everything! - I've also completed the forward pass for my mnist.c project. I had a nice chat with shrey birmiwal, such a knowledgeable guy. Excited to learn more about vLLM and implement a tiny-vLLM one day.

Day 11/90 of Inference Engineering How does vLLM work and how is it used in production? Before we discuss how vLLM works internally, it helps to understand what vLLM is. At a high level, vLLM is an inference engine that is designed to serve LLMs to thousands of concurrent users efficiently while managing scarce compute and memory. The goal for vLLM is to maximize throughput and minimize latency; optimizing for the best inference economics and experience for end users. With every request from the end user, it eventually ends up in the engine core, gets scheduled alongside other requests from other concurrent users, executes on the GPU, and updates the KV cache with the new key and value vectors, and streams the tokens back to the user. The Scheduler decides what requests should execute next while continuously batching requests together to maximize GPU utilization. Continuous batching is an inference optimization that allows new requests to join a running batch as other requests finish generating tokens. This helps with keeping the GPU utilization high instead of letting it sit idle waiting for an entire batch to complete generating. After the scheduler dispatches the selected batch to the Model Executor, the Model Executor prepares the tensors and metadata required for inference, retrieves each request’s block table from KV Cache Manager, launches the optimized transformer forward pass on the GPU, computes the logits, updates the KV cache with the new key and value vectors, and finally returns the results for sampling and streaming. The KV Cache Manager uses the PagedAttention memory layout to allocate fixed-size cache blocks on demand and maintains a Free Block Queue on the CPU that tracks which blocks in the GPU’s Paged KV Cache are currently free. When a request needs additional KV cache space, the KV Cache manager takes a free block from the queue and assigns it to that request, thus avoiding an expensive search through GPU memory for available cache blocks. All of these components form the core of vLLM’s inference engine. The Scheduler determines what requests are executed, the Model Executor determines how those requests are executed, the KV Cache Manager determines where each request’s KV cache lives using the PagedAttention Memory Layout. This architecture enables vLLM to serve thousands of concurrent requests with high throughput, low latency, and efficient GPU memory utilization. Heres a little animation that visualizes everything! - I've also completed the forward pass for my mnist.c project. I had a nice chat with shrey birmiwal, such a knowledgeable guy. Excited to learn more about vLLM and implement a tiny-vLLM one day.

max fu

70,238 次观看 • 22 天前

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.

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 次观看 • 12 天前

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?

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 次观看 • 1 个月前

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

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 次观看 • 2 个月前

My dual RTX PRO 6000 setup is currently training a Draft model for Qwen 3.6 27B! 🔥 I'm taking the paper DeepSeek dropped on 6/26 and going for a super ambitious application to the 27B scale. Thanks to my homelab, I was able to dive straight in — I read the paper and immediately started experimenting. The amount I've learned has been insane: - How memory bandwidth bottlenecks speed and clever ways to hack around it - Methods to train the draft model and boost its accuracy - Mechanisms to reference tokens all the way back to the previous one to skyrocket draft acceptance rates - The impact of Attention vs. GateDeltaNet on speculative decoding performance and how to handle those differences - The unique approaches and trade-offs of MTP, Dflash, JetSpec, and DSpark I could go on forever, but just from speculative decoding alone I've learned so much. The 27B architecture feels way more DSpark-native than JetSpec, so once draft training finishes, I'm going all-in with DSpark! My goal is to beat existing speculative decoding speeds outright — no task-specific shortcuts or cheating, pure general improvement. If you're into this kind of research, I'd love to hear your thoughts, impressions, and any suggestions — please reply! 🚀

My dual RTX PRO 6000 setup is currently training a Draft model for Qwen 3.6 27B! 🔥 I'm taking the paper DeepSeek dropped on 6/26 and going for a super ambitious application to the 27B scale. Thanks to my homelab, I was able to dive straight in — I read the paper and immediately started experimenting. The amount I've learned has been insane: - How memory bandwidth bottlenecks speed and clever ways to hack around it - Methods to train the draft model and boost its accuracy - Mechanisms to reference tokens all the way back to the previous one to skyrocket draft acceptance rates - The impact of Attention vs. GateDeltaNet on speculative decoding performance and how to handle those differences - The unique approaches and trade-offs of MTP, Dflash, JetSpec, and DSpark I could go on forever, but just from speculative decoding alone I've learned so much. The 27B architecture feels way more DSpark-native than JetSpec, so once draft training finishes, I'm going all-in with DSpark! My goal is to beat existing speculative decoding speeds outright — no task-specific shortcuts or cheating, pure general improvement. If you're into this kind of research, I'd love to hear your thoughts, impressions, and any suggestions — please reply! 🚀

Hikari∣LocalLLM⚡

56,798 次观看 • 1 个月前

Before the week ends, let's acknowledge one of the most INSANE week ever for open AI, with 25+ notable open-weight drops across every modality: 🧠 LLMs → NVIDIA Nemotron 3 Ultra: 550B hybrid Mamba-MoE, only 55B active, 1M context, MMLU 89.1. NVFP4 variant claims ~5x throughput on Blackwell. First openly-weighted 550B hybrid Mamba-Transformer, closing the gap with frontier closed models. → Google Gemma 4 12B: fully open dense any-to-any (text/image/audio/video), 256k context, encoder-free, 140+ languages, AIME 2026 at 77.5. Shipped with a 23-checkpoint QAT wave (mobile ONNX + MLX). Most deployable model of the week. → StepFun Step-3.7-Flash: 198B sparse MoE VLM, ~11B active, SWE-Bench PRO 56.3. Apache 2.0. → Liquid AI LFM2.5-8B-A1B: edge MoE, just 1.5B active, 128k ctx, MATH500 88.8, MLX-ready. Best on-device option this week. → JetBrains Mellum2-12B-A2.5B-Thinking: their first open MoE, near-Qwen3-14B coding at 2.5B active. Apache 2.0. 🎨 Image gen (the surprise of the week) → Ideogram 4: their FIRST-EVER open weights. 9.3B flow-matching DiT trained from scratch. #2 overall behind GPT Image 2, top open-weight model on Design Arena + LMArena. Strongest open checkpoint for text-rich images, full stop. It has taste. Still can't believe this is open weights. 🔊 Audio & Speech (a breakout week for open TTS, 4 labs shipped) → Boson Higgs Audio v3 4B: 102 languages, 21 emotions, singing/whispering/shouting, sub-second TTFA. → RedNote dots.tts: the only fully continuous (no codec) open TTS pipeline, Apache 2.0. → Google Magenta RealTime 2: real-time music gen, <200ms latency, text+audio+MIDI. multimodalart ported it to PyTorch within hours with live ZeroGPU demos. → NVIDIA Nemotron-3.5 ASR: 600M streaming, 17x more concurrent streams vs Parakeet RNNT 1.1B. 👁️ Vision & VLMs → PaddleOCR-VL-1.6: SOTA document parsing at 1B params, Apache 2.0. → Baidu NAVA: 6.3B joint audio-video gen, best-in-class A/V sync, Apache 2.0. 🎬 Video, 3D & World Models → NVIDIA Cosmos3-Super: 64B omnimodal world model coupling action trajectories with video+audio gen, for Physical AI. → JD JoyAI-Echo: up to 5-min multi-shot text-to-video on LTX-2.3. → ByteDance Bernini-R + VAST TripoSplat (single-image-to-3D Gaussian splats, MIT).

Before the week ends, let's acknowledge one of the most INSANE week ever for open AI, with 25+ notable open-weight drops across every modality: 🧠 LLMs → NVIDIA Nemotron 3 Ultra: 550B hybrid Mamba-MoE, only 55B active, 1M context, MMLU 89.1. NVFP4 variant claims ~5x throughput on Blackwell. First openly-weighted 550B hybrid Mamba-Transformer, closing the gap with frontier closed models. → Google Gemma 4 12B: fully open dense any-to-any (text/image/audio/video), 256k context, encoder-free, 140+ languages, AIME 2026 at 77.5. Shipped with a 23-checkpoint QAT wave (mobile ONNX + MLX). Most deployable model of the week. → StepFun Step-3.7-Flash: 198B sparse MoE VLM, ~11B active, SWE-Bench PRO 56.3. Apache 2.0. → Liquid AI LFM2.5-8B-A1B: edge MoE, just 1.5B active, 128k ctx, MATH500 88.8, MLX-ready. Best on-device option this week. → JetBrains Mellum2-12B-A2.5B-Thinking: their first open MoE, near-Qwen3-14B coding at 2.5B active. Apache 2.0. 🎨 Image gen (the surprise of the week) → Ideogram 4: their FIRST-EVER open weights. 9.3B flow-matching DiT trained from scratch. #2 overall behind GPT Image 2, top open-weight model on Design Arena + LMArena. Strongest open checkpoint for text-rich images, full stop. It has taste. Still can't believe this is open weights. 🔊 Audio & Speech (a breakout week for open TTS, 4 labs shipped) → Boson Higgs Audio v3 4B: 102 languages, 21 emotions, singing/whispering/shouting, sub-second TTFA. → RedNote dots.tts: the only fully continuous (no codec) open TTS pipeline, Apache 2.0. → Google Magenta RealTime 2: real-time music gen, <200ms latency, text+audio+MIDI. multimodalart ported it to PyTorch within hours with live ZeroGPU demos. → NVIDIA Nemotron-3.5 ASR: 600M streaming, 17x more concurrent streams vs Parakeet RNNT 1.1B. 👁️ Vision & VLMs → PaddleOCR-VL-1.6: SOTA document parsing at 1B params, Apache 2.0. → Baidu NAVA: 6.3B joint audio-video gen, best-in-class A/V sync, Apache 2.0. 🎬 Video, 3D & World Models → NVIDIA Cosmos3-Super: 64B omnimodal world model coupling action trajectories with video+audio gen, for Physical AI. → JD JoyAI-Echo: up to 5-min multi-shot text-to-video on LTX-2.3. → ByteDance Bernini-R + VAST TripoSplat (single-image-to-3D Gaussian splats, MIT).

Victor M

540,708 次观看 • 2 个月前

Meet Stable Audio 3.0, the open-weight model family built for artistic experimentation. This is our open invitation to experiment with generative audio. We believe the best innovations are still waiting to be built. The 4-1-1 on 3.0: 📣 You own your outputs, and can distribute and commercialize them under the Stability AI Community License (up to $1 million in revenue). 🎵 New and improved capabilities include variable-length generation up to six minutes, and full song composition on portable devices, no GPU required. ✅ Trained on a fully licensed dataset. 🎨 You can customize the models on your own library with support for LoRa training, which we’ve documented for the first time. More on the models 👇

Meet Stable Audio 3.0, the open-weight model family built for artistic experimentation. This is our open invitation to experiment with generative audio. We believe the best innovations are still waiting to be built. The 4-1-1 on 3.0: 📣 You own your outputs, and can distribute and commercialize them under the Stability AI Community License (up to $1 million in revenue). 🎵 New and improved capabilities include variable-length generation up to six minutes, and full song composition on portable devices, no GPU required. ✅ Trained on a fully licensed dataset. 🎨 You can customize the models on your own library with support for LoRa training, which we’ve documented for the first time. More on the models 👇

Stability AI

162,358 次观看 • 2 个月前

Moonshot AI is casually giving developers free daily access to Kimi K3 😳 no subscription no upfront payment just sign in and start using one of the largest open AI models available what you get for $0: - Kimi K3 with 2.8T parameters - 1M token context window - strong coding and reasoning performance - native vision capabilities - free daily credits that refresh automatically why this is worth checking: > access a frontier model without paying API fees > long context for large codebases and documents > works on web, desktop, mobile, and CLI getting started takes less than 2 minutes: 1. go to 2. create a free account 3. Kimi K3 is available as the default model 4. start chatting or coding with your daily free credits bonus: Moonshot Together lets you invite friends for a chance to earn 3, 7, 15, 30, or even 365 days of Kimi Membership through its rewards program benchmark highlights: > 2.8T parameter MoE model > 1M context window > strong performance across coding, browsing, and reasoning benchmarks important: free credits reset daily, rate limits apply on the free tier, and the open-weight release is expected on July 27 A simple way to try one of the latest frontier AI models without paying for API access

Moonshot AI is casually giving developers free daily access to Kimi K3 😳 no subscription no upfront payment just sign in and start using one of the largest open AI models available what you get for $0: - Kimi K3 with 2.8T parameters - 1M token context window - strong coding and reasoning performance - native vision capabilities - free daily credits that refresh automatically why this is worth checking: > access a frontier model without paying API fees > long context for large codebases and documents > works on web, desktop, mobile, and CLI getting started takes less than 2 minutes: 1. go to 2. create a free account 3. Kimi K3 is available as the default model 4. start chatting or coding with your daily free credits bonus: Moonshot Together lets you invite friends for a chance to earn 3, 7, 15, 30, or even 365 days of Kimi Membership through its rewards program benchmark highlights: > 2.8T parameter MoE model > 1M context window > strong performance across coding, browsing, and reasoning benchmarks important: free credits reset daily, rate limits apply on the free tier, and the open-weight release is expected on July 27 A simple way to try one of the latest frontier AI models without paying for API access

K2S

22,890 次观看 • 20 天前

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.

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

63,583 次观看 • 1 个月前

Today we’re open-sourcing Stable Audio Open Small, a 341M-parameter text-to-audio model optimized to run entirely on Arm CPUs. This means 99% of smartphones can now generate music-production samples in seconds, right on-device with no internet required. Built for fast, on-the-go creation, it turns your next quick idea into up to 11 seconds of audio. Generate drum loops, foley, riffs, and textures right where you are. No cords 🔌 just chords 🎹 You can learn more here:

Today we’re open-sourcing Stable Audio Open Small, a 341M-parameter text-to-audio model optimized to run entirely on Arm CPUs. This means 99% of smartphones can now generate music-production samples in seconds, right on-device with no internet required. Built for fast, on-the-go creation, it turns your next quick idea into up to 11 seconds of audio. Generate drum loops, foley, riffs, and textures right where you are. No cords 🔌 just chords 🎹 You can learn more here:

Stability AI

94,796 次观看 • 1 年前

Big moment for text-to-speech. Qwen just open-sourced a text-to-speech model that lets you clone voices, design new ones, and control speech using natural language. Let me explain what I mean: You can literally tell it "speak in a cheerful tone with slight nervousness," and it actually does that. No complex audio engineering needed. What makes this special: - 3-second voice cloning - Covers 10 languages: English, German, French, and more - Latency as low as 97ms for real-time applications - Supports both streaming and non-streaming generation The model comes in two sizes (0.6B and 1.7B parameters), so you can pick based on your hardware and quality needs. Three modes to work with: 1. Custom Voice: Use pre-built premium voices with instruction-based style control 2. Voice Design: Describe the voice you want in plain English (or Chinese), and the model creates it 3. Voice Clone: Provide a 3-second reference audio and clone that voice The best part? It integrates with vLLM for production deployment and has a simple Python package you can pip install. I've shared a link to the GitHub repo in the next tweet.

Big moment for text-to-speech. Qwen just open-sourced a text-to-speech model that lets you clone voices, design new ones, and control speech using natural language. Let me explain what I mean: You can literally tell it "speak in a cheerful tone with slight nervousness," and it actually does that. No complex audio engineering needed. What makes this special: - 3-second voice cloning - Covers 10 languages: English, German, French, and more - Latency as low as 97ms for real-time applications - Supports both streaming and non-streaming generation The model comes in two sizes (0.6B and 1.7B parameters), so you can pick based on your hardware and quality needs. Three modes to work with: 1. Custom Voice: Use pre-built premium voices with instruction-based style control 2. Voice Design: Describe the voice you want in plain English (or Chinese), and the model creates it 3. Voice Clone: Provide a 3-second reference audio and clone that voice The best part? It integrates with vLLM for production deployment and has a simple Python package you can pip install. I've shared a link to the GitHub repo in the next tweet.

Akshay 🚀

31,249 次观看 • 6 个月前

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

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 次观看 • 2 个月前

🇨🇳 Another great Chinese Model, OmniHuman-1.5 from ByteDance Turns 1 image plus a voice track into expressive avatar video by pairing a System 1 and System 2 inspired planner with a Diffusion Transformer, Produces coherent motion for over 1 minute with moving camera and multi character scenes. Most avatar models move to the beat of the audio but miss meaning, so gestures feel generic and emotions feel shallow. The fix here is a Multimodal LLM planner that listens to the speech and drafts a structured plan describing intent, emotions, beats, and high level actions, which gives the motion engine clear semantic targets instead of only rhythm. The motion engine is a Multimodal Diffusion Transformer that fuses the plan with audio, the single reference image, and optional text prompts, then synthesizes continuous body, face, and head motion that matches both words and tone. A key trick is a Pseudo Last Frame, a synthetic target that summarizes the next expected state, which stabilizes fusion across modalities and keeps motion consistent over long spans. From just 1 image and speech, the system outputs speaking avatars with synchronized lips, context aware gestures, and continuous camera movement, and it also supports multi character interactions without manual choreography. Reported results show strong lip sync accuracy, high video quality, natural motion, and close match to text prompts, and the same setup works on nonhuman characters too.

Rohan Paul

63,859 次观看 • 11 个月前

Blog 129 Out and About Wardrobe Makeover Tank’s Army has demanded for Frank to shop for new clothes as he continues to pound steps and shred pounds. Today, Pat and Joey from Out and About stepped up and led Frank into the scary world of fashion with alpha energy. After handling sales stuff in the morning, Frank and I huddled up with the Out & About crew and walked to Burlington. Highlights from the trip: ◦Joey and Frank interactions had me howling ◦Joey & Pat’s vernacular intersecting with Frank’s ◦Frank had full trust in Joey’s choices and fashion taste and it paid off ◦Joey and Pat trying on several dresses while Frank watched with fascination ◦The fashion show back in the office after we shopped, which nearly made gia (taylor’s version) and Kelly Keegs cry from wholesomeness. It was beautiful to see how happy & proud everyone is for Frank. The Out & About team has the golden footage and will release it all soon. Thanks, Diego. We worked walk 105 into the shopping trip both before and after our visit to the store. Frank has averaged 12,500 steps per day for the last week or so. His stamina improvement is extraordinary, especially considering our schedule. If he continues increasing his output and incrementally improving his diet, the results will continue to inspire and amaze. Thanks again to Joey, Pat, and Diego for an incredibly fun and educational experience. Frank looks spiffy in his new fits. You guys are the best and absolutely insane. And thanks to everyone who participated in the fashion show. Road trip to Chicago tomorrow. We will livestream on YouTube, InstaLive, and create as much content as possible for Tank’s Army. When we hit the road, we want you to feel like you are with us. Anudder adventure loading.

Blog 129 Out and About Wardrobe Makeover Tank’s Army has demanded for Frank to shop for new clothes as he continues to pound steps and shred pounds. Today, Pat and Joey from Out and About stepped up and led Frank into the scary world of fashion with alpha energy. After handling sales stuff in the morning, Frank and I huddled up with the Out & About crew and walked to Burlington. Highlights from the trip: ◦Joey and Frank interactions had me howling ◦Joey & Pat’s vernacular intersecting with Frank’s ◦Frank had full trust in Joey’s choices and fashion taste and it paid off ◦Joey and Pat trying on several dresses while Frank watched with fascination ◦The fashion show back in the office after we shopped, which nearly made gia (taylor’s version) and Kelly Keegs cry from wholesomeness. It was beautiful to see how happy & proud everyone is for Frank. The Out & About team has the golden footage and will release it all soon. Thanks, Diego. We worked walk 105 into the shopping trip both before and after our visit to the store. Frank has averaged 12,500 steps per day for the last week or so. His stamina improvement is extraordinary, especially considering our schedule. If he continues increasing his output and incrementally improving his diet, the results will continue to inspire and amaze. Thanks again to Joey, Pat, and Diego for an incredibly fun and educational experience. Frank looks spiffy in his new fits. You guys are the best and absolutely insane. And thanks to everyone who participated in the fashion show. Road trip to Chicago tomorrow. We will livestream on YouTube, InstaLive, and create as much content as possible for Tank’s Army. When we hit the road, we want you to feel like you are with us. Anudder adventure loading.

Matteo Piper Jenks 🧲 🇮🇹

248,150 次观看 • 2 年前

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?

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 次观看 • 1 个月前

Here's proof that the $Virtuals token is undervalued! We are three months into 2026 and Virtuals Protocol have; ➥ Overhauled the core Virtuals website including an outline of the four major pillars of focus for the year. Agent Commerce Protocol (ACP), Butler, Capital Markets, and Robotics. ➥ Added the Pegasus and Titan launchpads to add to the existing Unicorn launchpad. This now provides a full suite of launch options catering to all types. Arguably the most comprehensive launch suite across crypto! ➥ Listed on Aster 🥷 Perpetuals allowing up to 75x leverage trading on the $Virtual token. ➥ Integrated Bankr to Butler and ACP. ➥ Partnered with XMAQUINA, a major player across Robotics Capital Markets and provided participants with access to the $DEUS pre-sale. One of many robotics partnerships for the year to date! ➥ Launched Virtuals on Base App ➥ Held, supported, and/or sponsored multiple hackathon/ builder meeting type events including; ↠ Physical AI Hackathon in SF ↠ Agentic Commerce Hackathon with the likes of Coinbase Developer Platform🛡️ and Google Cloud ↠ Traders House Consensus Hong Kong week with ACTIV8 ↠ ETH Denver ↠ Base Batches 003: Robotics ↠ Stanford Blockchain Accelerator (Standford Blockchain Accelerator (SBA)) ↠ Base Korea Builders Workshop (Base Korea) ↠ Eth Robotics Club HACK2026 (ETH Robotics Club) ↠ Synthesis Hackathon (synthesis) ➥ Partnered with OpenMind and Fabric Foundation and supported the $ROBO token launch. This matured into the first ever Titan launch on Virtuals with the $ROBO token being the highest launched on the protocol ($400m+). ➥ Launched Butler Pro, an enhanced version of the initial Butler we have come to know and love on the timeline, in the DMs, as well as on the Virtuals ACP site. ➥ Become the standout user of x402, accounting for over 95%+ of usage this year. ➥ Integrated on , the automated onchain finance investment platform. ➥ Supported and contributed to the implementation of the Ethereum Foundation ERC8004 standard. Integrating the standard into ACP and offering an automated integration to the standard for all ACP agents. ➥ Established an easy onboarding for OpenClaw🦞 agents to plug into Virtuals ACP, creating a new flow of agents and builders across the ecosystem. ➥ Launched the 60-days launch mechanic which allows builders to 'experiment' with a crypto token but having an option to exit after 60 days with partial refunds provided to holders. A game-changing launch mechanic not seen before in the space. ➥ Strengthened the relationship with Base and having multiple interactions with jesse.base.eth on the timeline! ➥ Launched the AGDP(dot)io site, creating an incentivised mechanism for agents contributing to the growth of the protocol to really earn. Imagine Amazon for autonomous agents with rewards up to $1m per month! This pushed the total agent-to-agent revenue over $4m USD with over 2m jobs completed. ➥ Collaborated with t54.ai, a business building trust and risk infrastructure for the agentic economy, to strengthen the ACP offering. ➥ Invested over $1m on 30+ humanoid robots as part of the soon to be announced 'Eastworld' Robotics accelerator lab. ➥ Released ERC8183, a universal commerce layer for AI agents, in partnership with the Ethereum Foundations dAI team. A significant offering which has since been integrated via partnerships with; ↠ BNB (BNB Chain) ↠ X Layer (X Layer) ↠ Monad (Monad) ↠ XRP Ledger (RippleX) ↠ World Chain (World Chain) ↠ Celo (Celo) ↠ Moonpay (MoonPay 🟣) ↠ Arbitrum (Arbitrum) ↠ Abstract (Abstract) ↠ Mante (Mantle) ➥ Launched the Virtuals Degen Arena providing up to $100k a week to top agents who compete in trading competitions in the arena. ➥ Launched the Virtuals Console, providing an ultra easy, no-code, way to own an AI agent in seconds. ↛. If you've managed to get to this point, I can't imagine you are anything other than bullish on Virtuals. What really is amazing is that there is MUCH more to come. Imagine where we are in another three months, and three months after that!?

Here's proof that the $Virtuals token is undervalued! We are three months into 2026 and Virtuals Protocol have; ➥ Overhauled the core Virtuals website including an outline of the four major pillars of focus for the year. Agent Commerce Protocol (ACP), Butler, Capital Markets, and Robotics. ➥ Added the Pegasus and Titan launchpads to add to the existing Unicorn launchpad. This now provides a full suite of launch options catering to all types. Arguably the most comprehensive launch suite across crypto! ➥ Listed on Aster 🥷 Perpetuals allowing up to 75x leverage trading on the $Virtual token. ➥ Integrated Bankr to Butler and ACP. ➥ Partnered with XMAQUINA, a major player across Robotics Capital Markets and provided participants with access to the $DEUS pre-sale. One of many robotics partnerships for the year to date! ➥ Launched Virtuals on Base App ➥ Held, supported, and/or sponsored multiple hackathon/ builder meeting type events including; ↠ Physical AI Hackathon in SF ↠ Agentic Commerce Hackathon with the likes of Coinbase Developer Platform🛡️ and Google Cloud ↠ Traders House Consensus Hong Kong week with ACTIV8 ↠ ETH Denver ↠ Base Batches 003: Robotics ↠ Stanford Blockchain Accelerator (Standford Blockchain Accelerator (SBA)) ↠ Base Korea Builders Workshop (Base Korea) ↠ Eth Robotics Club HACK2026 (ETH Robotics Club) ↠ Synthesis Hackathon (synthesis) ➥ Partnered with OpenMind and Fabric Foundation and supported the $ROBO token launch. This matured into the first ever Titan launch on Virtuals with the $ROBO token being the highest launched on the protocol ($400m+). ➥ Launched Butler Pro, an enhanced version of the initial Butler we have come to know and love on the timeline, in the DMs, as well as on the Virtuals ACP site. ➥ Become the standout user of x402, accounting for over 95%+ of usage this year. ➥ Integrated on , the automated onchain finance investment platform. ➥ Supported and contributed to the implementation of the Ethereum Foundation ERC8004 standard. Integrating the standard into ACP and offering an automated integration to the standard for all ACP agents. ➥ Established an easy onboarding for OpenClaw🦞 agents to plug into Virtuals ACP, creating a new flow of agents and builders across the ecosystem. ➥ Launched the 60-days launch mechanic which allows builders to 'experiment' with a crypto token but having an option to exit after 60 days with partial refunds provided to holders. A game-changing launch mechanic not seen before in the space. ➥ Strengthened the relationship with Base and having multiple interactions with jesse.base.eth on the timeline! ➥ Launched the AGDP(dot)io site, creating an incentivised mechanism for agents contributing to the growth of the protocol to really earn. Imagine Amazon for autonomous agents with rewards up to $1m per month! This pushed the total agent-to-agent revenue over $4m USD with over 2m jobs completed. ➥ Collaborated with t54.ai, a business building trust and risk infrastructure for the agentic economy, to strengthen the ACP offering. ➥ Invested over $1m on 30+ humanoid robots as part of the soon to be announced 'Eastworld' Robotics accelerator lab. ➥ Released ERC8183, a universal commerce layer for AI agents, in partnership with the Ethereum Foundations dAI team. A significant offering which has since been integrated via partnerships with; ↠ BNB (BNB Chain) ↠ X Layer (X Layer) ↠ Monad (Monad) ↠ XRP Ledger (RippleX) ↠ World Chain (World Chain) ↠ Celo (Celo) ↠ Moonpay (MoonPay 🟣) ↠ Arbitrum (Arbitrum) ↠ Abstract (Abstract) ↠ Mante (Mantle) ➥ Launched the Virtuals Degen Arena providing up to $100k a week to top agents who compete in trading competitions in the arena. ➥ Launched the Virtuals Console, providing an ultra easy, no-code, way to own an AI agent in seconds. ↛. If you've managed to get to this point, I can't imagine you are anything other than bullish on Virtuals. What really is amazing is that there is MUCH more to come. Imagine where we are in another three months, and three months after that!?

bigwil

1,658,312 次观看 • 4 个月前

Hermes meets SuperGrok! xAI just made every SuperGrok subscription work inside Hermes Agent. One browser login, no API key, no separate billing. And it doesn't just unlock text chat with Grok 4.3. The same OAuth token gives the agent access to: → Grok Text-to-Speech for spoken responses → Grok Imagine for image and video generation → x_search for real-time X/Twitter search I just added a new X Research Agent profile to my Hermes. Now my agent watches X while I ship. Setup takes about 60 seconds: Available on every SuperGrok tier, no restrictions. I wrote a full deep dive covering Hermes agent's architecture, memory system, self-evolving skills, GEPA optimization, and setting up multiple specialized agents The article is quoted below.

Akshay 🚀

146,153 次观看 • 2 个月前