ModelScope just put a countdown on Qwen3.8-Flash-Next. No weights... yet. Card says multimodal MoE, 125B total, 6B active, plus a 51B n-gram embedding. GDN + QSA. They call it the Qwen4 architecture, shipping early so people can prep. I already run Qwen 3.8-27B on Spark 1. This is a different job. Wait for files on Qwen/Qwen3.8-Flash-Next before you grab a GGUF. Estimated drop: Aug 26, 11am ET.show more

Qwen 3.8 Flash Next (125B A6B MoE) Vs Qwen... 3.8 27B (dense) - Both Q4_K_XL - Reasoning Off Prompt: Create a very high quality realistic video like animation of the solar system. three.js via cdn. Let the camera revolve around. single html file. dont call any tools. dont need any controls. Qwen 3.8 27b focused on computational physics & geometry (inclinations, coordinate math, moon orbits). While Qwen 3.8 125B A6B focused on cinematography & visual UX (lighting contrast, atmospheric glow, framing, and HUD overlays). It also followed the camera instruction correctly and produced result thats visually more stunning. Both the models run on single RTX 3090/4090 (24 GB VRAM). the 27b will fit entirely in the VRAM, but you'd need 100-120GB RAM for the new 125B MoE (check out the previous post for complete inference benchmarks and llama.cpp setup) Qwen 3.8 Flash Next beats Qwen 3.8 27B in almost all the benchmarks. have dropped the intelligence benchmark comparison and unsloth quant huggingface links in the replies. Which one would u be running regularly on your hardware?show more

QWEN 3.8 MAX + DEEPSEEK V4 PRO BOTH FREE... ON ONE ENDPOINT • the gateway > Qwen3.8-Max (2.4T MoE) + DeepSeek V4 Pro free tiers: > Sign up -> one API key -> start with no card for free tier. > Free is tight by design -- they’re letting you taste the real thing before you scale. • wiring it up > Base url: >Model IDs exact: qwen/qwen3.8-max-free and deepseek/deepseek-v4-pro-0813-free. > Drop straight into Cursor/Claude Code/any OpenAI-compatible client. Good for long agent runs you leave open on a second screen, slow for anything interactive. Point it at side projects and experiments, not at the code you would mind a stranger reading ↓show more

Qwen 3.8 27B Q4_K_M - 90 tokens/sec on a... single NVIDIA RTX 4090 (24 GB VRAM) with Dflash2! (MTP 60 tps -> 90 tps Dflash2!!!!) Local AI moves so fast (literally!) it’s terrifying. Z lab just dropped DFlash 2 for Qwen 3.8 27b and Muse Glimmer. I patched llama.cpp (PR #27342) and paired it with Unsloth’s Qwen 3.8 27B UD-Q4_K_XL quant. The result? Lossless 90 tokens/s decode. My last post highlighted native MTP hitting 60 t/s at 130,000 context. But DFlash 2 just completely shattered that ceiling. By using parallel block diffusion drafting (predicting whole blocks of tokens in a single pass using dynamic convolutions), DFlash achieves a massive 5.39 token acceptance rate. THE ALPHA TWEAK: `n-max 7` eats too much VRAM for draft states. But if you drop the draft limit to `--spec-draft-n-max 4`, you slash the VRAM overhead and actually increase the throughput. Here is the new 24GB VRAM Physics Matrix (DFlash 2 @ n-max 4): - 30k Context: 1,725 t/s prefill | 87.05 t/s decode | 22.2 GB VRAM - 80k Context: 1,789 t/s prefill | 84.20 t/s decode | 23.3 GB VRAM - 110k Context: 1,767 t/s prefill | 83.35 t/s decode | 23.96 GB VRAM (110k context at 83+ tokens a second sitting exactly on the 24GB hardware limit is absolute wizardry). How to compile the PR today: git clone cd llama.cpp git fetch origin pull/27342/head:pr-27342 git switch pr-27342 cmake -B build -DGGML_CUDA=ON && cmake --build build -j Llama.cpp flags for Dflash (110k Context Ceiling): ./build/bin/llama-server -m Qwen3.8-27B-UD-Q4_K_XL.gguf -md Qwen3.8-27B-DFlash2-Q4_K_M.gguf --spec-type draft-dflash --spec-draft-n-max 4 -c 110000 -ngl 99 --port 8080 -ctv q4_0 -ctk q4_0 The fact that the open source community is shipping block diffusion drafters so quickly that run entirely locally on a gaming GPU is unbelievable. If you own a single RTX 3090 or 4090, it is officially time to upgrade to qwen 3.8 27b with dflash 2 and cancel your API subscriptions and let local silicon eat the cloud. This model beats GPT 5.6 Terra, GLM 5.2 DeepSeek V4 Pro, Muse Spark 1.2 and Claude Opus 4.8 on the artificial analysis agentic index (details in the replies) Hugging Face GGUF links (Base + DFlash2) and the full visual VRAM scaling and Dflash2 vs MTP graphs are also in the replies below. are you sticking to native MTP for the 130k context, or sacrificing 20k context to redline your decode speed? How many tokens/sec are you pushing on your current local rig?show more

CT always asks where the edge is. It’s not... charts. It’s not indicators. It’s not copy trading. It’s being early. Same way people made fortunes grabbing $ASTEROID early. Same way we positioned on $SI before it moved. Now we’re here, live, hearing the next narrative get named: Superior Intelligence. No media coverage yet. No clips yet (besides this one, badly recorded by me ). Pure insidoor play , approx 200 people were in mar a lago attending Trumps speaking about building not an Artificial Intelligence empire, but a SUPERIOR INTELLIGENCE one. No noise. Just one spark. (A public Trump’s interview, an Elon Musk comment, an Sam Altman quote, a Mark Zuckerberg mention… and it flies.) And that’s all it takes for SI to moon. I already see , NVIDIA , Meta , Claude , Grok posting about SUPERIOR INTELLIGENCE One narrative → one ignition → full explosion. If you wait for confirmation, you lose your hedge ! Front run the next 100 M narrative potential on eth 0x91139d452c144486D49B388509e459AB8748d07Bshow more

Qwen3.8-Flash-Next is starting to feel like the local model... Opus fans have been waiting for. Someone ran the NVFP4 176B-class Flash-Next on 2× DGX Sparks, and the results are wild. Real measured scaling → C1: 44.2 tok/s → C2: 64.6 tok/s → C4: 86.8 tok/s aggregate The per-stream speed drops with concurrency, but total throughput keeps climbing. Long-context behavior was even more impressive: → 5K: needle retrieved → 21K: needle retrieved → 84K: needle retrieved → 167K: needle retrieved → 262K: prefill succeeded, but the window was saturated That 167K retrieval test is the one I care about. Long agent runs are where models usually start losing the plot. Flash-Next didn’t. It also held up surprisingly well on physics-heavy reasoning, artifact generation, research workflows, evidence checking, and long-horizon planning. The personality is interesting too. DeepSeek V4 Flash feels like the dependable workhorse. GLM-5.2 feels like the problem-solving machine. Qwen3.8-Flash-Next feels more insightful. It has that rare ability to understand what you’re actually asking rather than just following the surface pattern. The main weakness I’ve noticed is instruction following. It can occasionally drift between prose turns where DeepSeek and GLM stay tighter. And this is why the 256GB M5 Ultra conversation gets interesting. If Apple can pair that huge unified-memory pool with enough bandwidth, this model class becomes genuinely practical for long-running local agents. We’re talking about frontier-class reasoning on hardware sitting on a desk.show more

THREE 3090s ON ONE BOARD GIVE YOU 72GB OF... VRAM AND KILL YOUR $200 CLAUDE CODE AND $200 OPENAI BILL people are pulling three used 3090s off ebay for around $2,100 total and stacking them in one tower to build a dedicated ai rig. that pools 72gb of vram for less than what a single rtx 5090 retails for alibaba shipped qwen 3.6 27b in april under apache 2.0. on realworldqa vision it scores 84.1 against claude 4.5 opus at 77.0. on ifbench instructions it lands at 76.5 against claude's 58.0 a single 3090 already runs qwen 3.6 27b with eight gigs of headroom. three of them in parallel handle larger models like deepseek r1 70b and qwen 235b without breaking a sweat a heavy ai user pays $200 claude code, $200 chatgpt pro plus $40 cursor and gemini. that's $5,280 a year and the rig pays itself off before month nine on $8 a month in electricity setup is one shell command for ollama, one to pull the model, one environment variable to point claude code at localhost. cli stays identical, nothing leaves the network, requests stop costing money bookmark this and read the article belowshow more

Throughout my journey in developing multimodal models, I’ve always... wanted a framework that lets me plug & play modality encoders/decoders on top of an auto-regressive LLM. I want to prototype fast, try new architectures, and have my demo files scale effortlessly — with full support for parallelism and optimization. Not just to hack⚙️, but also to scale🚀. So finally we built it for ourselves. LMMs-Engine: a lean, efficient framework built to train unified multimodal model at scale. From Qwen LLM, VLM, LLaVA-OV, and WanVideo, to unified models like Qwen-Omni and BAGEL — plus Linear-Attn GDN and research prototypes like RAE and SiT - all under one modular system that seamlessly integrates diverse datasets and optimization strategies. Powered by FSDP2 multi-dim parallelism, Ulysses sequence parallel, Flash-Attention, Liger Kernels, and Native Sparse Attention (also with bonus support for the Muon optimizer for all models).show more

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.show more

5 days ago it took 2 GPUs to build... this. today it takes 1. same prompt. same particle simulation. completely different model. Qwen-Coder-Next (80B) on 2x 3090s. 46 tok/s. 564 lines. 2 iterations to get it working. 48GB VRAM across two cards just to hold it. Qwen3.5-35B-A3B on a single 3090. 112 tok/s. 461 lines. first try. cleaner code, fewer lines, better structured. 19.7GB on disk with 4GB VRAM to spare. half the parameters. one GPU instead of two. 2.4x faster. and the output actually improved. this is what happens when architecture catches up to ambition. Gated Delta Networks(Mamba2 variant) hybrid with sparse MoE. 3B active params out of 35B per token. efficiency at the architecture level, not just quantization. the curve isn't flattening. it's steepening.show more

Qwen 3.8 27B at 56tps; on 9 year old... GPU btw Nvidia V100 32GB ~$650 on EBay right now! Using Dflash 2; disabling the ECC adds some more speed too! Thinking and prose is a bit slower, but 56-63 tps in code gen! MTP runs faster for prose vs DFlash2 but slower sustained code generation speed. MTP also runs much faster power limited than DFlash does. Working on a repo so you can get up and going quickly. Fun fact, the Nvidia v100 was $11,500 per card when they first launched. Price you pay for future proofing I guess; they’re still great cards. Pcie 3.0 and the older software/architecture are the only drawbacks, but also those aren’t as much of an issue as you’d think. Especially when you consider the price today!show more

🚨 BREAKING: An American teenager just used the new... Qwen 3.6 Plus model to unlock an infinite money glitch on Polymarket. While Qwen 3.6 is currently smoking Claude, this kid used it to build a psychopathic trading bot dedicated entirely to 5-minute Bitcoin candles. Operating under the username nickwwwo, this script isn't analyzing macro trends or fed rates. It's predicting the literal next 300 seconds of BTC price action while the rest of the world sleeps. The stats are genuinely terrifying: ///> Total PnL: +$30,800 ///> Win Rate: 61.7% (on 5-minute binary options, this is basically time travel) ///> Total Gains: $55,018 --- Just look at the receipts: At 2:35 AM, the bot placed a $120 bet that BTC would drop. Exactly five minutes later, it cashed out $5,773. That is a 4,686% ROI in the time it takes to brush your teeth. Right now, half the timeline is trying to manually track his wallet to catch his entries. But on a 5-minute timeframe, if you're trading by hand, you're already exit liquidity. The play isn't to beat the bot, it's to clone it. 🧬 You can fully automate this and copytrade his exact moves using the ARES. The millisecond the nickwwwo algorithm enters a position, Ares mirrors it for your account. Try copying this wallet with ares: Stop fighting the machines, and let them print for you.show more

50% more context unlocked for Qwen 3.8 27b Q4_K_XL... dflash 2 on a single RTX 4090 (24 GB VRAM) I found a hidden VRAM tax in llama.cpp. By combining my custom 2 bit DFlash 2 drafter with one overlooked server flag, I just unlocked another +80,000 tokens of context. Qwen3.8-27B is now running a massive 250,000 context at 75 tokens/s on a single RTX 4090. Here is the secret: By default, `llama-server` reserves massive chunks of your VRAM to handle multiple concurrent users (batching). If you are running a single user session, you are bleeding memory for features you aren't using. By passing the `--parallel 1` flag, you force the engine to dedicate 100% of your 24GB VRAM buffer to a single user. When we combine the VRAM saved by our Q2_K 2-bit drafter with the VRAM saved by `--parallel 1`, the context ceilings absolutely explode: Note: all benchmarks carried out with a massive 28k prompt. Ubuntu 22. ### THE NEW 24GB PHYSICAL LIMITS (Single RTX 4090): # 1. The "Repo Swallower" (Q4 KV Cache): - Context: 250,000 tokens (Up from 170k!) - Speed: 73.66 t/s decode | 1,608 t/s prefill - Peak VRAM: 23.8 GB # 2. The "High-Precision SWE" (Q8 KV Cache): - Context: 150,000 tokens (Up from 100k!) - Speed: 75.01 t/s decode | 1,667 t/s prefill - Peak VRAM: 23.9 GB # 3. The "Pristine Attention" (Unquantized FP16 KV): - Context: 90,000 tokens - Speed: 80.58 t/s decode | 1,699 t/s prefill - Peak VRAM: 23.92 GB ### HOW TO RUN THE 250K GOD STACK TODAY: (Requires PR #27342 + my Q2_K Hugging Face drafter) llama.cpp flags: ./build/bin/llama-server -m Qwen3.8-27B-UD-Q4_K_XL.gguf -md Qwen3.8-27B-DFlash2-Q2_K.gguf --spec-type draft-dflash --spec-draft-n-max 3 -c 250000 -ngl 99 --parallel 1 --port 8080 -ctv q4_0 -ctk q4_0 We are pushing a quarter million tokens of context with speculative DFlash 2 decoding at 73 tokens/second on a single consumer gaming GPU. I dropped my custom 2 bit Hugging Face GGUF links, visual performance graphs, and the PR #27342 build instructions in the replies below. If you own a single RTX 3090 or 4090, it is officially time to cancel your API subscriptions and let local silicon eat the cloud. how much monthly API spend does an optimized 4090 rig like this actually replace for you?show more

Affiliates are making $12,000 a month from AI reaction... clips right now and barely anyone is talking about it No camera. No face. No filming. → Pick a gift card offer on Glitchy → Reverse engineer a viral reaction clip in Claude → Generate your version in Sora → Post across 10 accounts targeting USA I collected 500 hooks people were already using to run these, fed them all into Claude, and got it to build me a better version of every single one You don't reinvent the wheel. You just build a faster one. It’s honestly the most slept on opportunity right now I put my whole process in a google doc If you want it RT this + comment “UGC” and I’ll send you it (must be following so I can DM)show more

Most people think their AI is not smart enough.... It is smart enough. It just knows nothing about you. Every chat starts from zero. You paste the doc. You explain the project. You explain the same project again tomorrow. I finally fixed that part: 👇 I have been running Littlebird for a while now. The idea is simple. It is a desktop app for Mac and Windows that reads the text on your active window and sits in on your calls. So it already knows what you have been working on before you ask it anything. You do not brief it. You just ask. What that looks like in a normal day: 1/ Chat that already has the context No pasting. I ask what changed in a brief last week and it answers from the actual document that was open on my screen. 2/ Meeting Notes that write themselves It transcribes the call, then hands me the decisions and the action items. I get to stay in the conversation instead of typing through it. 3/ Routines that run on a schedule A morning briefing. A weekly summary of what I actually shipped. It shows up on its own. 4/ Hummingbird for the small stuff It appears right where you are working, so a quick question does not cost you a window switch and ten minutes of drift. The point is not the notes. Plenty of apps take notes. The point is that I stopped re-explaining my own work to a machine fifteen times a day. Free plan if you want to test the idea before deciding. Link is in the first comment, along with a discount for new users.show more

Inkling-small is out today! With SGLang, you can get... 648 tok/s decode with DSpark (simulated acc len=4) and 288 tok/s w/o DSpark, under the same setup (8x NVIDIA AI B200, TP 8, NVFP4, bs=1). What makes this model different is the size. 276B total with 12B active is a sweet spot for RL, and both LoRA and full-parameter training become well within reach. Miles is ready and verified for multimodal RL on Inkling-small, so you can turn your multimodal data into real capability gains. At ~1/4 the size, Inkling-small matches the bigger version in capability and even wins on some benchmarks. Run Inkling-small with SGLang, and customize it with Miles.show more

🤮 15 seconds of pure 7-Eleven nightmare fuel just... dropped… You grab a soda or Slurpee, take a couple sips, decide “nah I’m good”… and according to this video they just wash the cup with hand soap and put it right back on the shelf for the next customer. “Nasty MOFOs.” Whether this is one rogue store or a sign of deeper corner-cutting, it’s the kind of thing that makes people stop going altogether. Meanwhile 7-Eleven just announced another massive wave of closures — 645 stores this year. No wonder.show more

The "I don't have enough VRAM" excuse just died.... I’m running Meta’s new 30B Muse Glimmer Q6_K_XL with a massive 130k context window on just 26GB VRAM FREE compute on Kaggle. Kaggle provides you free 2x Nvidia T4 GPUs. 30 hours usage each week! Yesterday, I showed you the violent throughput of Muse Glimmer on a single RTX 4090. Today, we are securing a Dual NVIDIA T4 GPU cluster with 32GB of total VRAM for exactly $0 and dropping the massive 24.5GB Q6_K_XL GGUF onto it. Here is the exact Kaggle workflow and benchmarking breakdown: # 1. The Storage Bypass & Setup I built a clean cell by cell script in the file. We dynamically fetch the CUDA accelerated llama.cpp binaries and use wget to stream the model directly into Kaggle's /kaggle/tmp scratch storage, which cleanly bypasses their 19.5GB output directory limit. # 2. The Multi GPU Performance With the -ngl 99 flag offloading all model layers across both T4 GPUs (32GB VRAM combined), we pushed a massive 131,072 token context window (-c 131072). The benchmark numbers: Prefill: 265.9 t/s Decode: 9.0 t/s VRAM Total: 26.5 GB # 3. The Architecture Insight The Q6_K_XL model itself is 24.5 GB. Because of Muse Glimmer's aggressive 16:1 GQA, the unquantized KV cache for a massive 130k context window only takes up 2 GB of memory. No heavily degraded Q4 KV quantization required. It just works. No compiling from source. No credit card. No OOM crashes. Zero excuses. If you’re running a single RTX 3090, 4090, or 5090, you need to experience this hyper efficient KV cache right now before the upcoming Qwen 3.8 27B drop completely steals your VRAM tomorrow. pick the Q4 or Q5 quants for 24 GB VRAM rigs. I'm dropping the Unsloth huggingface GGUF links and the free Kaggle notebook link in the replies. spin up your own instance, and show me your multi GPU benchmarks.show more

"But where's the flash?" Watch this CNN demonstration of... 6g of PETN being ignited by open flame. What happens? The PETN burns with a visible flame for several seconds. Then it detonates. The detonation happens so fast there is no visible flash — the camera goes straight from fire to debris field. And here's the key: the fire that WAS there is blown OUT by the blast wave. This is not a shaped charge. This is unconfined PETN in open air. No flash. No fireball. The blast wave actually extinguishes the existing flame. Why? Three reasons: PETN's reaction zone is measured in microns and completes in nanoseconds (Anderson et al., Propellants Explosives Pyrotechnics, 2022). At gram scale, the entire detonation event is over in single-digit microseconds. A 30fps camera captures 33,000 μs per frame. The event occupies <0.03% of one frame. The visible "flash" people expect from explosions comes from compression-heating of surrounding air — not the explosive itself. At gram scale there simply isn't enough gas volume being heated to produce visible light that registers on a standard camera. In a shaped charge, it's even less visible because the energy is directed INTO the target as a hydrodynamic metal jet (Munroe effect), not radiated outward as heat and light. The Hezbollah pager attacks (Sept 2024) used 3-6g PETN per device. Watch the CCTV footage — no fireballs. Just a pop and casualties. Sandia National Labs detonates ~32mg PETN and researchers stand next to the chamber in safety glasses. No flash. No fire. "No flash = no explosive" is a Hollywood education, not a physics one. Joe Rogan shawshow more

I just built a Meta Ads diagnostic in Claude... Code that tells you WHY your account broke, not just what changed 🤯 It spins up a team of agents that each investigate a different reason performance dropped, then argue against each other to kill the wrong answer before it ever reaches you. All inside Claude Code. Perfect for DTC brands and agencies who panic-kill creative the second CPA spikes. If you've watched ROAS fall off a cliff and opened Ads Manager with ten tabs going, you already know what happens next. Your gut says "creative fatigue." You kill your best-performing ad. A week later performance is still broken, because that was never the problem. Guessing wrong is the most expensive move in paid social. This workflow ends the guessing: → One agent investigates each competing theory — creative fatigue, budget and delivery changes, traffic quality, offer and seasonality → Each one is blind to the others, reasoning only from its own slice of the data so they can't bias each other → A refuter agent then attacks every surviving theory and tries to kill it → A theory only stands if the data can't disprove it → You get a ranked diagnosis: the real cause, the evidence for and against it, and the one move to make this week No anchoring on the first obvious answer. No killing winning creative on a hunch. No "here's what happened" reports that never tell you why. What you get: → Every theory tested in parallel instead of one biased guess → An adversarial pass that kills the wrong answer before you act on it → A ranked diagnosis with confidence levels and evidence both ways → A reusable workflow you drop next month's export into and re-run Built 100% in Claude Code with the new dynamic workflows. The first account I ran it on looked like textbook creative fatigue. The workflow disagreed, and traced the real cause to a budget change that had doubled spend and flooded delivery with junk traffic. I put together a full playbook with the exact workflow, the prompt, and how to run it on your own account. Want it for free? > Like this post > Comment "META" And I'll send it over (must be following so I can DM)show more

clipping agencies rent offices now, $500,000 a year ten... editors, a manager, a lease, and the only thing any of them is paid for is noticing noticing is 310 lines of python, and i gave it away it runs on Robinhood Chain, so the clip pays whoever cut it CUT never sees a single frame of the stream. it reads chat, because chat knows before the editor does, before the dashboard does, before the streamer does [what those 310 lines do] 1. measure the last 10 seconds against that stream's own 5 minute baseline 2. never a global number, a big channel idling still beats a small one peaking 3. count how many people just typed clip it 4. call the clip api in the same second, not after the room reacts that last one is the whole thing. a twitch clip only keeps the 90 seconds before you hit the button, so if you noticed the moment yourself, you already lost it no key to read chat. it cannot post for you. it runs on your machine, not mine one command replays three minutes of recorded chat offline and fires a real moment on your screen, so you can watch it work before you trust itshow more