Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

hy4 preview changes the question from “can a lab run 770B?” to “can your team stand it up?” tencent hunyuan released an open w8 text model: 𝟕𝟕𝟎𝐁 𝐭𝐨𝐭𝐚𝐥, 𝟒𝟗𝐁 𝐚𝐜𝐭𝐢𝐯𝐞 𝐩𝐞𝐫 𝐭𝐨𝐤𝐞𝐧, Apache 2.0 and native 𝟏𝐦 𝐜𝐨𝐧𝐭𝐞𝐱𝐭. Day 0 paths exist for vLLM and SGLang, plus an official...

future_coded's profile picture

Future Coded

26,131 subscribers

32,855 Aufrufe • vor 24 Tagen •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

11 Kommentare

Profilbild von Liam | AI Tools & News
Liam | AI Tools & Newsvor 23 Tagen

Amazing

Profilbild von Max
Maxvor 24 Tagen

Really enjoyed this one. There’s a thoughtful perspective here that genuinely stayed with me.

Profilbild von NOVA
NOVAvor 24 Tagen

214GiB mixed quant is the part that actually matters

Profilbild von Iris Hayes
Iris Hayesvor 23 Tagen

Apache 2.0 plus 1M context is a serious ownership move

Profilbild von Liam
Liamvor 23 Tagen

Patched llama.cpp will slow early testers

Profilbild von Amber Nexus
Amber Nexusvor 23 Tagen

Masterpiece

Profilbild von Vikas gupta
Vikas guptavor 23 Tagen

Need real coding and long-context numbers, not just size claims

Profilbild von Rachel Woods
Rachel Woodsvor 24 Tagen

Still not a laptop model no matter how you slice 214GiB

Profilbild von Brian Hadu
Brian Haduvor 24 Tagen

if your team can't handle it, deploying 770B won't matter at all

Profilbild von Muhammad Ali
Muhammad Alivor 24 Tagen

49B active is what makes this class even serveable

Profilbild von Shahid Ansari
Shahid Ansarivor 24 Tagen

Tokens per second on your own box is the only score that counts.

Ähnliche Videos

What can a 770B model actually build? Hy4 preview is making that question more practical. 770B total / 49B active Native 1M context Apache 2.0 Day-0 vLLM + SGLang Official FP8 ~1.5TB → ~214GiB with mixed per-layer quant The interesting part: reported accuracy barely moves vs BF16. Here’s Hy4 turning a simple prompt into a playable game. A 770B-class model with a much lower deployment barrier. Tencent Hy
0:40

What can a 770B model actually build? Hy4 preview is making that question more practical. 770B total / 49B active Native 1M context Apache 2.0 Day-0 vLLM + SGLang Official FP8 ~1.5TB → ~214GiB with mixed per-layer quant The interesting part: reported accuracy barely moves vs BF16. Here’s Hy4 turning a simple prompt into a playable game. A 770B-class model with a much lower deployment barrier. Tencent Hy

Charles Roman

12,224 Aufrufe • vor 24 Tagen

Hy4 preview powers dynamic builds for a 770B open model. 770B total parameters, 49B active per token, native 1M context, Apache 2.0, direct vLLM and SGLang support, plus official FP8 weights. In this demo, Hy4 creates a colorful, playable 2D space shooter game directly from a prompt, featuring active controls, enemies, visual particle effects, plus score tracking. It is a preview. Test it, measure tokens/s, compare it with BF16 or FP8, and see what it builds. Try it in WorkBuddy: Tencent Hy WorkBuddy #Hy4Preview #OpenSourceAI #GameDev #AI
1:16

Hy4 preview powers dynamic builds for a 770B open model. 770B total parameters, 49B active per token, native 1M context, Apache 2.0, direct vLLM and SGLang support, plus official FP8 weights. In this demo, Hy4 creates a colorful, playable 2D space shooter game directly from a prompt, featuring active controls, enemies, visual particle effects, plus score tracking. It is a preview. Test it, measure tokens/s, compare it with BF16 or FP8, and see what it builds. Try it in WorkBuddy: Tencent Hy WorkBuddy #Hy4Preview #OpenSourceAI #GameDev #AI

Tyler Wayne

78,023 Aufrufe • vor 20 Tagen

Laguna S 2.1 is built to run locally. We heard the ask for a serious open model in the 80–120B range. Laguna S 2.1 is our answer, and we want it to redefine what this size class can do. Weights ship in BF16, FP8, INT4 and NVFP4, with official GGUF and MLX conversions, plus official DFlash draft models for faster inference. Day-one support across llama.cpp, ollama, Valerie Lee Make Up, SGLang, NVIDIA TRT-LLM, Hugging Face transformers and ZML. Watch it run on an NVIDIA AI DGX Spark.
0:39

Laguna S 2.1 is built to run locally. We heard the ask for a serious open model in the 80–120B range. Laguna S 2.1 is our answer, and we want it to redefine what this size class can do. Weights ship in BF16, FP8, INT4 and NVFP4, with official GGUF and MLX conversions, plus official DFlash draft models for faster inference. Day-one support across llama.cpp, ollama, Valerie Lee Make Up, SGLang, NVIDIA TRT-LLM, Hugging Face transformers and ZML. Watch it run on an NVIDIA AI DGX Spark.

Poolside

31,242 Aufrufe • vor 2 Monaten

The next big advantage in AI might not come from one breakthrough. It might come from how quickly a team can build, learn, and improve. Tencent Hunyuan has now released three major models in six months. Hy4 preview is the latest result of that pace. It is fully open source, with 770B total parameters, 49B active parameters, and support for more than 1M tokens of context. The model is designed around practical work, including coding, office productivity, science, planning, and longer workflows that require sustained execution. Access is priced at $0.834 per million input tokens, $2.501 per million output tokens, and $0.042 per million cache hits. That combination of rapid iteration, open access, and affordable usage is what makes this release worth paying attention to. Hy4 preview can be experienced free on WorkBuddy for the next two weeks. Tencent Hy Tencent AI WorkBuddy
0:39

The next big advantage in AI might not come from one breakthrough. It might come from how quickly a team can build, learn, and improve. Tencent Hunyuan has now released three major models in six months. Hy4 preview is the latest result of that pace. It is fully open source, with 770B total parameters, 49B active parameters, and support for more than 1M tokens of context. The model is designed around practical work, including coding, office productivity, science, planning, and longer workflows that require sustained execution. Access is priced at $0.834 per million input tokens, $2.501 per million output tokens, and $0.042 per million cache hits. That combination of rapid iteration, open access, and affordable usage is what makes this release worth paying attention to. Hy4 preview can be experienced free on WorkBuddy for the next two weeks. Tencent Hy Tencent AI WorkBuddy

Manish Kumar Shah

42,263 Aufrufe • vor 1 Monat

I’ve been testing Hy4 preview in WorkBuddy, and the most interesting part is not simply the model size, it’s how much practical work it can handle with a relatively focused active parameter count. Hy4 preview brings together stronger code understanding, generation, and editing; improved document and information processing; workflow automation; web and game development; cross-tool collaboration; and more reliable completion of complex, multi-step tasks. In other words, it is designed for work that requires planning, tool use, iteration, and follow-through, not just a quick answer in a chat window. Compared with its initial release, the current Hy4 preview is noticeably faster and better-performing in practical workflows. Following an upgrade released yesterday, it can complete tasks with fewer conversation rounds and lower token usage, while reasoning more quickly and making the overall user experience feel smoother from the first instruction to the final result. For my test, I gave it a demanding Three.js game-prototyping task with a 770B-parameter model and 49B active parameters. The result was more revealing than a simple first-look demo: Hy4 preview handled the core logic, edge cases, and follow-up changes while maintaining the broader context of the project. That combination of capability, speed, context, and active compute is what makes its cost-effectiveness worth examining. A fair evaluation should use the same prompt and environment configuration across models, changing only the model itself. That makes it easier to assess task completion, planning quality, tool-calling stability, reasoning speed, token efficiency, and performance over longer workflows without confusing the result with different settings. If you want to test the model yourself, access Hy4 preview through WorkBuddy and see how it performs on a real coding, document, automation, or creative task: Tencent Hy Tencent AI WorkBuddy
1:02

I’ve been testing Hy4 preview in WorkBuddy, and the most interesting part is not simply the model size, it’s how much practical work it can handle with a relatively focused active parameter count. Hy4 preview brings together stronger code understanding, generation, and editing; improved document and information processing; workflow automation; web and game development; cross-tool collaboration; and more reliable completion of complex, multi-step tasks. In other words, it is designed for work that requires planning, tool use, iteration, and follow-through, not just a quick answer in a chat window. Compared with its initial release, the current Hy4 preview is noticeably faster and better-performing in practical workflows. Following an upgrade released yesterday, it can complete tasks with fewer conversation rounds and lower token usage, while reasoning more quickly and making the overall user experience feel smoother from the first instruction to the final result. For my test, I gave it a demanding Three.js game-prototyping task with a 770B-parameter model and 49B active parameters. The result was more revealing than a simple first-look demo: Hy4 preview handled the core logic, edge cases, and follow-up changes while maintaining the broader context of the project. That combination of capability, speed, context, and active compute is what makes its cost-effectiveness worth examining. A fair evaluation should use the same prompt and environment configuration across models, changing only the model itself. That makes it easier to assess task completion, planning quality, tool-calling stability, reasoning speed, token efficiency, and performance over longer workflows without confusing the result with different settings. If you want to test the model yourself, access Hy4 preview through WorkBuddy and see how it performs on a real coding, document, automation, or creative task: Tencent Hy Tencent AI WorkBuddy

Tyler Wayne

56,152 Aufrufe • vor 19 Tagen

I tested Tencent’s Hy4 preview model in WorkBuddy on a real frontend build, not a benchmark screenshot. I gave it one practical brief: create an original neon courier game in a single HTML file, with Canvas rendering, keyboard controls, collision detection, scoring, a countdown timer, a boost mechanic, sound effects, and reliable restart logic. The point was not to see whether it could describe a game. I wanted to see whether it could turn a creative idea into a coherent, playable result while handling the details that often break a quick prototype: state changes, movement, collisions, feedback, layout, and interaction flow. Tencent’s Hy4 preview is designed for stronger code understanding, generation, and editing, together with document and information processing, workflow automation, web and game development, cross-tool collaboration, and complex multi-step task completion. That makes it relevant not only for answering questions, but also for taking a task from brief to working output. The current preview version is faster and more effective than the first release. After the latest upgrade, the workflow takes fewer conversation rounds, uses fewer tokens, and reaches useful results more quickly. The difference is most noticeable in the handoff between idea, implementation, revision, and final testing: there is less waiting and less back-and-forth before the result becomes usable. For context, the reference materials describe Tencent’s Hy4 preview as a 770B-parameter model with 49B active parameters, a 1M-token context window, and Apache 2.0 weights. The practical question is how those capabilities translate into real work. In this case, I’m testing whether the model can produce a finished, playable game rather than just a promising code fragment. The one-minute video shows the complete process: selecting Tencent’s Hy4 preview model in WorkBuddy, submitting the build prompt, reviewing the generated result, playing the game, and checking the controls, collisions, score, timer, boost effect, sound, and restart flow. Try your own coding, frontend, document, or automation task in WorkBuddy: Tencent Hy Workbuddy Tencent AI
0:31

I tested Tencent’s Hy4 preview model in WorkBuddy on a real frontend build, not a benchmark screenshot. I gave it one practical brief: create an original neon courier game in a single HTML file, with Canvas rendering, keyboard controls, collision detection, scoring, a countdown timer, a boost mechanic, sound effects, and reliable restart logic. The point was not to see whether it could describe a game. I wanted to see whether it could turn a creative idea into a coherent, playable result while handling the details that often break a quick prototype: state changes, movement, collisions, feedback, layout, and interaction flow. Tencent’s Hy4 preview is designed for stronger code understanding, generation, and editing, together with document and information processing, workflow automation, web and game development, cross-tool collaboration, and complex multi-step task completion. That makes it relevant not only for answering questions, but also for taking a task from brief to working output. The current preview version is faster and more effective than the first release. After the latest upgrade, the workflow takes fewer conversation rounds, uses fewer tokens, and reaches useful results more quickly. The difference is most noticeable in the handoff between idea, implementation, revision, and final testing: there is less waiting and less back-and-forth before the result becomes usable. For context, the reference materials describe Tencent’s Hy4 preview as a 770B-parameter model with 49B active parameters, a 1M-token context window, and Apache 2.0 weights. The practical question is how those capabilities translate into real work. In this case, I’m testing whether the model can produce a finished, playable game rather than just a promising code fragment. The one-minute video shows the complete process: selecting Tencent’s Hy4 preview model in WorkBuddy, submitting the build prompt, reviewing the generated result, playing the game, and checking the controls, collisions, score, timer, boost effect, sound, and restart flow. Try your own coding, frontend, document, or automation task in WorkBuddy: Tencent Hy Workbuddy Tencent AI

Rebecca Adson

61,820 Aufrufe • vor 12 Tagen

Alibaba just released a coding model that hits 82 percent on SWE-Bench Verified. That is the highest score ever published for an open-source model. The weights are free. The license is Apache 2.0. You can run it today. The model is Qwen 4 Coder 32B. Here is what 82 percent on SWE-Bench Verified actually means. SWE-Bench Verified tests whether an AI can autonomously resolve real bugs pulled from real production GitHub repositories. Not synthetic exercises. Real open-source projects that real teams depend on. A model gets a bug report, reads the code, writes a fix, and either passes the test suite or it does not. At 82 percent, Qwen 4 Coder 32B resolves 82 out of every 100 real production bugs it is given. Without a human guiding it. On code it has never seen before. For comparison: Qwen 4 Coder 32B: 82 percent SWE-Bench Verified. Open source. Apache 2.0. Claude Fable 5: 80.3 percent SWE-Bench Pro. $10 input / $50 output per million tokens. Currently suspended. GPT-5.6 Sol: Competitive on Terminal-Bench. $5 input / $30 output per million tokens. An open-weight model that you can download and run for free just beat both of them on the benchmark designed to measure real software engineering capability. Here is the architecture. Qwen 4 Coder 32B is a 32 billion parameter dense model. Not a Mixture-of-Experts. Every parameter is active on every request. This matters for inference: a dense 32B model runs on 22 gigabytes of VRAM, which fits on a single high-end consumer GPU or a MacBook Pro with 64GB of unified memory. The smaller variant, Qwen 4 Coder 4B, runs at approximately 135 tokens per second on an M5 Max and fits inside 8 gigabytes of RAM. For a model with usable coding capability, that is a new bar for what fits in a single laptop. The training methodology continued Alibaba's approach of reinforcement learning on verifiable coding tasks. The model gets rewarded when its code passes tests. It gets penalized when it fails. Over millions of training steps, the model learns to write code that actually runs rather than code that looks plausible. License: Apache 2.0. Full commercial use. No attribution requirement. No revenue threshold. No monthly active user ceiling. Weights: Hugging Face, available today. Runs on: vLLM, Ollama, SGLang, and any standard GGUF-compatible inference engine. Qwen 4 32B also runs at approximately 135 tokens per second on an M5 Max chip, setting a new bar for what a sub-8GB model can do on Apple Silicon. The open-source coding model just beat the best closed-source model in the world on the benchmark designed to test whether AI can actually do software engineering. The weights are free. The subscription is optional. Source: Autom8Labs AI Insight July 2026, State of Open Source LLMs June 2026, Kunal Ganglani blog June 2026.
0:25

Alibaba just released a coding model that hits 82 percent on SWE-Bench Verified. That is the highest score ever published for an open-source model. The weights are free. The license is Apache 2.0. You can run it today. The model is Qwen 4 Coder 32B. Here is what 82 percent on SWE-Bench Verified actually means. SWE-Bench Verified tests whether an AI can autonomously resolve real bugs pulled from real production GitHub repositories. Not synthetic exercises. Real open-source projects that real teams depend on. A model gets a bug report, reads the code, writes a fix, and either passes the test suite or it does not. At 82 percent, Qwen 4 Coder 32B resolves 82 out of every 100 real production bugs it is given. Without a human guiding it. On code it has never seen before. For comparison: Qwen 4 Coder 32B: 82 percent SWE-Bench Verified. Open source. Apache 2.0. Claude Fable 5: 80.3 percent SWE-Bench Pro. $10 input / $50 output per million tokens. Currently suspended. GPT-5.6 Sol: Competitive on Terminal-Bench. $5 input / $30 output per million tokens. An open-weight model that you can download and run for free just beat both of them on the benchmark designed to measure real software engineering capability. Here is the architecture. Qwen 4 Coder 32B is a 32 billion parameter dense model. Not a Mixture-of-Experts. Every parameter is active on every request. This matters for inference: a dense 32B model runs on 22 gigabytes of VRAM, which fits on a single high-end consumer GPU or a MacBook Pro with 64GB of unified memory. The smaller variant, Qwen 4 Coder 4B, runs at approximately 135 tokens per second on an M5 Max and fits inside 8 gigabytes of RAM. For a model with usable coding capability, that is a new bar for what fits in a single laptop. The training methodology continued Alibaba's approach of reinforcement learning on verifiable coding tasks. The model gets rewarded when its code passes tests. It gets penalized when it fails. Over millions of training steps, the model learns to write code that actually runs rather than code that looks plausible. License: Apache 2.0. Full commercial use. No attribution requirement. No revenue threshold. No monthly active user ceiling. Weights: Hugging Face, available today. Runs on: vLLM, Ollama, SGLang, and any standard GGUF-compatible inference engine. Qwen 4 32B also runs at approximately 135 tokens per second on an M5 Max chip, setting a new bar for what a sub-8GB model can do on Apple Silicon. The open-source coding model just beat the best closed-source model in the world on the benchmark designed to test whether AI can actually do software engineering. The weights are free. The subscription is optional. Source: Autom8Labs AI Insight July 2026, State of Open Source LLMs June 2026, Kunal Ganglani blog June 2026.

Harman

41,278 Aufrufe • vor 2 Monaten

I gave an AI one instruction: “Build me a playable 3D game from scratch.” Then I let it cook. 👀 I tested Tencent Hy4 preview inside WorkBuddy, and it went way beyond just generating a few snippets of code. It could: → Plan the game structure → Generate the code across files → Create the UI and interactions → Debug problems along the way → Iterate on the project while keeping context That last part is what caught my attention. Hy4 preview is built for these long-horizon tasks, with 770B total parameters, 49B active parameters, and a 1M-token context window. It’s designed for coding agents, game development, tool use, research, and other workflows where the model has to actually do the work, not just answer a prompt. And right now, you can try Hy4 preview for free on WorkBuddy until September 10. If you want to see how far an AI agent can take a real project, try it yourself: See what you can build with it. Tencent AI
1:27

I gave an AI one instruction: “Build me a playable 3D game from scratch.” Then I let it cook. 👀 I tested Tencent Hy4 preview inside WorkBuddy, and it went way beyond just generating a few snippets of code. It could: → Plan the game structure → Generate the code across files → Create the UI and interactions → Debug problems along the way → Iterate on the project while keeping context That last part is what caught my attention. Hy4 preview is built for these long-horizon tasks, with 770B total parameters, 49B active parameters, and a 1M-token context window. It’s designed for coding agents, game development, tool use, research, and other workflows where the model has to actually do the work, not just answer a prompt. And right now, you can try Hy4 preview for free on WorkBuddy until September 10. If you want to see how far an AI agent can take a real project, try it yourself: See what you can build with it. Tencent AI

Toha Khan

66,214 Aufrufe • vor 25 Tagen

your agent reviewing its own work is not a check. it is a second opinion from the same source. this is the most common gap in agent systems and it hides in plain sight, because the step exists. there is a review. it just cannot do the thing you think it does. here is the mechanism. the model produced an output from a context. you then ask the same model, holding the same context, whether that output is correct. it answers fluently, because that is what it does. and the answer is drawn from the same distribution that produced the thing being judged. same weights, same window, same blind spots. if the reason the output is wrong is something the model does not know, the review does not know it either. if the reason is something the context does not contain, the review has the same context. the failure mode and the detector share a cause. > why it feels like it works because most of the time the output is fine, and the review says fine. agreement is not evidence of detection. a reviewer that says pass on everything agrees with reality most of the time too. what you actually want to measure is what happens on the cases that are wrong. that is the only place a check earns its name, and it is exactly the place where a self-review is weakest. there is research on this. Huang and colleagues at DeepMind showed at ICLR 2024 that intrinsic self-correction, revising without external grounding, does not reliably help and often makes things worse. > what to actually do move the check outside the model. a test that runs, a schema that validates, a file that exists or does not, an exit code from something you did not write. these are not smarter than the model. they are just not correlated with it, and that is the entire value. when the judgement genuinely needs a model, at minimum use a different family. same family means shared blind spots, and frontier judges measurably inflate scores for outputs that look like their own. and split the work by kind. anything objectively checkable goes to code. only the genuinely semantic calls go to a judge, and those get a rubric written as one line. a review inside the loop tells you the model is confident. a check outside it tells you whether the work is done. save this - then read the eval setup below
0:50

your agent reviewing its own work is not a check. it is a second opinion from the same source. this is the most common gap in agent systems and it hides in plain sight, because the step exists. there is a review. it just cannot do the thing you think it does. here is the mechanism. the model produced an output from a context. you then ask the same model, holding the same context, whether that output is correct. it answers fluently, because that is what it does. and the answer is drawn from the same distribution that produced the thing being judged. same weights, same window, same blind spots. if the reason the output is wrong is something the model does not know, the review does not know it either. if the reason is something the context does not contain, the review has the same context. the failure mode and the detector share a cause. > why it feels like it works because most of the time the output is fine, and the review says fine. agreement is not evidence of detection. a reviewer that says pass on everything agrees with reality most of the time too. what you actually want to measure is what happens on the cases that are wrong. that is the only place a check earns its name, and it is exactly the place where a self-review is weakest. there is research on this. Huang and colleagues at DeepMind showed at ICLR 2024 that intrinsic self-correction, revising without external grounding, does not reliably help and often makes things worse. > what to actually do move the check outside the model. a test that runs, a schema that validates, a file that exists or does not, an exit code from something you did not write. these are not smarter than the model. they are just not correlated with it, and that is the entire value. when the judgement genuinely needs a model, at minimum use a different family. same family means shared blind spots, and frontier judges measurably inflate scores for outputs that look like their own. and split the work by kind. anything objectively checkable goes to code. only the genuinely semantic calls go to a judge, and those get a rubric written as one line. a review inside the loop tells you the model is confident. a check outside it tells you whether the work is done. save this - then read the eval setup below

Hanako

14,325 Aufrufe • vor 1 Monat

America is about to lose the AI race, and it will not happen at the frontier. It will happen at the floor. Everyone is watching who ships the smartest model. The actual war is over the 80% of tokens nobody posts about: the routine inference that quietly runs the world. On current trajectory, that fight is already lost. Watch what people do, not what they say. Coinbase just defaulted its own engineers off frontier models onto open weights and cut AI spend nearly in half while usage kept climbing. Even NVIDIA runs a closed frontier model as an orchestrator and pushes the volume to its own open weights. The frontier is becoming a router. The volume goes open. That part is settled. Here is the part that should terrify Washington: the only credible open tier today is Chinese. GLM. Kimi. And the US answer is to tighten export controls and freeze its own labs in place, as if you can embargo a file that is already downloaded, or price-match free. So China hands the Global South Huawei hardware and free open models, and a generation in Africa and Southeast Asia learns to reason through a model that will not tell them what happened at Tiananmen Square. That is not a cost story. That is influence through inference. You do not have to win hearts and minds when you supply the mind. Open source is not a nice-to-have for America. It is the whole ballgame for the 80%, and right now the US is barely on the field. We need American open weights. Not eventually. Now.
0:47

America is about to lose the AI race, and it will not happen at the frontier. It will happen at the floor. Everyone is watching who ships the smartest model. The actual war is over the 80% of tokens nobody posts about: the routine inference that quietly runs the world. On current trajectory, that fight is already lost. Watch what people do, not what they say. Coinbase just defaulted its own engineers off frontier models onto open weights and cut AI spend nearly in half while usage kept climbing. Even NVIDIA runs a closed frontier model as an orchestrator and pushes the volume to its own open weights. The frontier is becoming a router. The volume goes open. That part is settled. Here is the part that should terrify Washington: the only credible open tier today is Chinese. GLM. Kimi. And the US answer is to tighten export controls and freeze its own labs in place, as if you can embargo a file that is already downloaded, or price-match free. So China hands the Global South Huawei hardware and free open models, and a generation in Africa and Southeast Asia learns to reason through a model that will not tell them what happened at Tiananmen Square. That is not a cost story. That is influence through inference. You do not have to win hearts and minds when you supply the mind. Open source is not a nice-to-have for America. It is the whole ballgame for the 80%, and right now the US is barely on the field. We need American open weights. Not eventually. Now.

Ben Pouladian

71,722 Aufrufe • vor 3 Monaten

UC Berkeley just open-sourced FreeToken. (2–4x faster local LLM inference than Ollama) the results are wild: - Qwen3.6-35B on an 8GB GPU at 39.3 tokens/s - DeepSeek-V4-Flash 284B on a 32GB GPU at 22 tokens/s - GLM-5.2 753B on a 96GB GPU at 14.9 tokens/s a 35B model at 16-bit precision needs about 70GB just for its weights. even at 4 bits it is close to 18GB, and FreeToken serves it on an 8GB GPU. let me explain how: all three models mentioned above are Mixture-of-Experts, and that is what FreeToken takes advantage of. each layer holds hundreds of separate experts plus a small router that picks a few of them per token. Qwen3.6-35B activates roughly 3B of its 35B parameters per token. DeepSeek-V4-Flash picks 6 of 256 experts per layer, so 13B of its 284B run at a time. so compute was never the bottleneck. the weights a single step touches fit comfortably on a consumer GPU. every expert the router might pick still has to exist somewhere. they sit in system RAM, and the GPU keeps a cache of the ones the model has been using recently. so everything comes down to what happens when the router picks an expert that is not on the GPU. there are two ways to serve that miss: 1. copy it over PCIe and run it on the GPU 2. run it on the CPU, where it already lives both read from the same system memory, so they compete for one pool of bandwidth instead of adding to each other. existing engines pick one option and freeze it when the model loads. but routing changes on every token, so a fixed choice misses most of what the model asks for. FreeToken measures both bandwidths on your machine and splits each step's misses between the two paths in proportion. the GPU and CPU results then merge exactly, with no approximation. two machines with the same GPU can end up wanting opposite strategies, which I did not expect. a 5090 in a gaming desktop should push nearly everything over PCIe, while an 8GB laptop is better off computing most misses on the CPU. none of that is readable off a spec sheet, so the engine profiles it once per machine. the second half of the design is about agents. coding agents constantly rewrite their own history, and every edit normally forces thousands of tokens back through prefill. FreeToken saves its checkpoints at the exact boundaries agent frameworks cut on, so it only reprocesses the new part. its slowest first token stays under 44 seconds, while llama.cpp peaks at 232 and KTransformers at 946. it serves the OpenAI and Anthropic APIs under Apache 2.0, so Claude Code and Codex can point at it directly. releasing weights publicly decides who can download a model, not who can afford to run one. frontier open models keep shipping, and running them still assumes a rented cluster. meanwhile there are over a hundred million consumer machines with discrete GPUs sitting mostly idle. closing that gap was never a hardware problem, and work like this is what turns open weights into something you can actually use. paper: repo: almost every idea in this post, from why memory bandwidth decides the outcome to why moving weights costs more than computing on them, comes straight out of how a GPU is built. I wrote a detailed primer on that. the article is quoted below.
0:42

UC Berkeley just open-sourced FreeToken. (2–4x faster local LLM inference than Ollama) the results are wild: - Qwen3.6-35B on an 8GB GPU at 39.3 tokens/s - DeepSeek-V4-Flash 284B on a 32GB GPU at 22 tokens/s - GLM-5.2 753B on a 96GB GPU at 14.9 tokens/s a 35B model at 16-bit precision needs about 70GB just for its weights. even at 4 bits it is close to 18GB, and FreeToken serves it on an 8GB GPU. let me explain how: all three models mentioned above are Mixture-of-Experts, and that is what FreeToken takes advantage of. each layer holds hundreds of separate experts plus a small router that picks a few of them per token. Qwen3.6-35B activates roughly 3B of its 35B parameters per token. DeepSeek-V4-Flash picks 6 of 256 experts per layer, so 13B of its 284B run at a time. so compute was never the bottleneck. the weights a single step touches fit comfortably on a consumer GPU. every expert the router might pick still has to exist somewhere. they sit in system RAM, and the GPU keeps a cache of the ones the model has been using recently. so everything comes down to what happens when the router picks an expert that is not on the GPU. there are two ways to serve that miss: 1. copy it over PCIe and run it on the GPU 2. run it on the CPU, where it already lives both read from the same system memory, so they compete for one pool of bandwidth instead of adding to each other. existing engines pick one option and freeze it when the model loads. but routing changes on every token, so a fixed choice misses most of what the model asks for. FreeToken measures both bandwidths on your machine and splits each step's misses between the two paths in proportion. the GPU and CPU results then merge exactly, with no approximation. two machines with the same GPU can end up wanting opposite strategies, which I did not expect. a 5090 in a gaming desktop should push nearly everything over PCIe, while an 8GB laptop is better off computing most misses on the CPU. none of that is readable off a spec sheet, so the engine profiles it once per machine. the second half of the design is about agents. coding agents constantly rewrite their own history, and every edit normally forces thousands of tokens back through prefill. FreeToken saves its checkpoints at the exact boundaries agent frameworks cut on, so it only reprocesses the new part. its slowest first token stays under 44 seconds, while llama.cpp peaks at 232 and KTransformers at 946. it serves the OpenAI and Anthropic APIs under Apache 2.0, so Claude Code and Codex can point at it directly. releasing weights publicly decides who can download a model, not who can afford to run one. frontier open models keep shipping, and running them still assumes a rented cluster. meanwhile there are over a hundred million consumer machines with discrete GPUs sitting mostly idle. closing that gap was never a hardware problem, and work like this is what turns open weights into something you can actually use. paper: repo: almost every idea in this post, from why memory bandwidth decides the outcome to why moving weights costs more than computing on them, comes straight out of how a GPU is built. I wrote a detailed primer on that. the article is quoted below.

Akshay 🚀

343,270 Aufrufe • vor 1 Monat

Memory vs. Graphs, clearly explained! memory is great, and the ceiling arrives quietly: it stores what happened. it does not store what to do about it. six runs later your file has fifty lines, and the model reloads all of them before it does anything. Graph engineering fixes this by changing what memory is: not a place things are kept, but an edge that runs backwards. you need both, and here is the sentence that resolves the whole confusion: a store keeps what happened. an edge keeps what to do about it. ↳ a store grows with every run, and every line is reloaded before the next one ↳ an edge carries one derived rule, and the rule replaces the run that produced it Prompts → Context → Harness → Loops → Graphs the transcript goes away, the constraint stays. and the constraint is smaller, because "adapters preserve keyword args exactly" is four hundred tokens shorter than the run that proved it. the same four blocks work on anything you can cut into lanes. i pointed them at token launches on Robinhood Chain, open source, nothing leaves your terminal the trick is knowing what deserves to survive. an output is not memory. "ported the utils slice, green on first pass" tells the next run nothing it can act on. the rule you derived from it does. one thing to know before you scale it. what you write down is not what comes back. ↳ the root rules file and auto memory are re-injected from disk. they come back intact, every time ↳ path-scoped rules live in message history. they get summarized away and do not return until a matching file is read again so a rule that must persist cannot be path-scoped. move it to the root and pay the always-loaded cost, or accept that it is advisory in any long session. and the one that eats whole nights: a memory file that has never had a line deleted is not memory. it is a tax on every run you will ever make, and nobody reads it back. below i have quoted my full guide on graph engineering. it covers the three topologies, the verifier patterns, and where the gate should actually open. save this, and the repo that runs it is below ↓
0:45

Memory vs. Graphs, clearly explained! memory is great, and the ceiling arrives quietly: it stores what happened. it does not store what to do about it. six runs later your file has fifty lines, and the model reloads all of them before it does anything. Graph engineering fixes this by changing what memory is: not a place things are kept, but an edge that runs backwards. you need both, and here is the sentence that resolves the whole confusion: a store keeps what happened. an edge keeps what to do about it. ↳ a store grows with every run, and every line is reloaded before the next one ↳ an edge carries one derived rule, and the rule replaces the run that produced it Prompts → Context → Harness → Loops → Graphs the transcript goes away, the constraint stays. and the constraint is smaller, because "adapters preserve keyword args exactly" is four hundred tokens shorter than the run that proved it. the same four blocks work on anything you can cut into lanes. i pointed them at token launches on Robinhood Chain, open source, nothing leaves your terminal the trick is knowing what deserves to survive. an output is not memory. "ported the utils slice, green on first pass" tells the next run nothing it can act on. the rule you derived from it does. one thing to know before you scale it. what you write down is not what comes back. ↳ the root rules file and auto memory are re-injected from disk. they come back intact, every time ↳ path-scoped rules live in message history. they get summarized away and do not return until a matching file is read again so a rule that must persist cannot be path-scoped. move it to the root and pay the always-loaded cost, or accept that it is advisory in any long session. and the one that eats whole nights: a memory file that has never had a line deleted is not memory. it is a tax on every run you will ever make, and nobody reads it back. below i have quoted my full guide on graph engineering. it covers the three topologies, the verifier patterns, and where the gate should actually open. save this, and the repo that runs it is below ↓

Hanako

47,766 Aufrufe • vor 22 Tagen

Qwen3.8-27B running at full BF16 on a free Kaggle TPU is kind of ridiculous. No quantization. No tiny context window. No expensive GPU instance. Just Qwen3.8-27B running on a Kaggle TPU v5e-8. The reported numbers: ~130 tok/s decode ~10,000 tok/s prefill 262K context That prefill number is especially wild. You can throw a huge amount of code or context at the model and ingest it extremely quickly, while still getting around 130 tokens per second during generation. And because it’s running in full BF16, you’re not relying on an aggressive quant just to make the model fit. But the really interesting part isn’t even the raw throughput. You can expose it as an OpenAI-compatible endpoint. That means you can plug the model into tools that already understand OpenAI-style APIs. Claude Code. Codex. OpenCode. And other compatible clients. So the workflow becomes pretty simple: Spin up the Qwen3.8-27B endpoint on Kaggle. Point your coding tool at the API. And suddenly you have a 27B coding model sitting behind the same interface you’d normally use for hosted models. The 262K context is also a huge deal for agentic coding. Large repositories can fit into a single context. Long conversations don’t need to be constantly trimmed. And tools can feed much more information back to the model without hitting a tiny context ceiling. The fact that this can be built around a free TPU environment is what makes this especially interesting. We’re getting to a point where experimenting with serious open models doesn’t always require owning a $2,000 GPU or paying for a large cloud instance. Free compute + open weights + an OpenAI-compatible API + existing coding agents. That’s a pretty powerful combination. Qwen3.8-27B is already an interesting model. Running the full BF16 version at ~130 tok/s with 262K context on free Kaggle TPU compute makes it a lot more interesting.
0:48

Qwen3.8-27B running at full BF16 on a free Kaggle TPU is kind of ridiculous. No quantization. No tiny context window. No expensive GPU instance. Just Qwen3.8-27B running on a Kaggle TPU v5e-8. The reported numbers: ~130 tok/s decode ~10,000 tok/s prefill 262K context That prefill number is especially wild. You can throw a huge amount of code or context at the model and ingest it extremely quickly, while still getting around 130 tokens per second during generation. And because it’s running in full BF16, you’re not relying on an aggressive quant just to make the model fit. But the really interesting part isn’t even the raw throughput. You can expose it as an OpenAI-compatible endpoint. That means you can plug the model into tools that already understand OpenAI-style APIs. Claude Code. Codex. OpenCode. And other compatible clients. So the workflow becomes pretty simple: Spin up the Qwen3.8-27B endpoint on Kaggle. Point your coding tool at the API. And suddenly you have a 27B coding model sitting behind the same interface you’d normally use for hosted models. The 262K context is also a huge deal for agentic coding. Large repositories can fit into a single context. Long conversations don’t need to be constantly trimmed. And tools can feed much more information back to the model without hitting a tiny context ceiling. The fact that this can be built around a free TPU environment is what makes this especially interesting. We’re getting to a point where experimenting with serious open models doesn’t always require owning a $2,000 GPU or paying for a large cloud instance. Free compute + open weights + an OpenAI-compatible API + existing coding agents. That’s a pretty powerful combination. Qwen3.8-27B is already an interesting model. Running the full BF16 version at ~130 tok/s with 262K context on free Kaggle TPU compute makes it a lot more interesting.

FHILY👑

35,910 Aufrufe • vor 24 Tagen

The most downloaded AI on earth is now Chinese. Alibaba just gave away a model that matches Claude's flagship, and it literally runs on a $700 used graphics card. The Qwen models crossed 3 BILLION downloads in six months. Hugging Face counted 418 million downloads for Google this year, and 227 million for Meta. Alibaba cleared more than four times both of them combined. Then today it released Qwen3.8-27B under an Apache 2.0 license. The model has 27 billion parameters, native vision, and a 262,000 token context window. Developers are running it locally on 17 gigabytes of memory, on used cards that cost a few hundred dollars. Alibaba's own benchmark table claims it beats Opus 4.6 Max on computer use by 84.3 to 72.7, on mobile use by 81.9 to 62, and on visual math by 94.6 to 65.5. Those numbers come from the vendor and nobody has independently verified them yet, so treat them as a claim. But the generation over generation jumps are harder to wave away: On DeepSWE the score went from 13.3 to 42.2. On software engineering it went from 49.3 to 79.0. That happened in ONE release cycle. And Apache 2.0 means anyone can download the weights, modify them, build products on them, sell those products, and never pay or ask permission. It cannot be revoked. Once the file is on your drive it is yours permanently. 3 billion downloads means those files already sit on machines in every country on Earth. Alibaba could delete everything tomorrow and it would change nothing. Washington spent 4 years building an export control regime around chips, model weights, and entity lists. Every piece of it assumes a chokepoint exists somewhere. A fab, a shipment, a company that can be told no. But there is no chokepoint for a file that has already been copied three billion times. And the copying compounds. Hugging Face counted 151,448 models built on top of Qwen, which is 2.6x Meta's entire footprint and 4.7x the number of Llama repositories. New ones appear at roughly 200 a day. The report says Qwen has become "part of the default workflow for developers deciding what models to fine-tune and deploy." Alibaba is also pushing Qwen through its cloud into Southeast Asia and Africa, markets where American labs have almost no presence, and where a very large share of the next generation of developers will learn to build. Meta and Nvidia have both rushed out new open models in recent weeks. That is what a response looks like when you feel the floor move. And to be clear, these are download and derivative numbers, not usage numbers. ChatGPT and Claude cannot be downloaded at all, so they do not appear in this comparison. What the figures measure is what developers choose to build on top of, which is a different question from what consumers type into a box. That is also why it matters MORE. Consumer habits change in an afternoon. Infrastructure choices last a decade, because everything built on top has to be rewritten to undo them. The American labs are valued on an assumption that frontier intelligence stays scarce, expensive, and rented by the token. Alibaba just made a version of it free, permanent, and small enough to run on hardware people already own. You will not get an announcement when the software you use every day starts running on a Chinese model underneath. Go and count how many of the tools you rely on could be rebuilt on free weights this year.
0:39

The most downloaded AI on earth is now Chinese. Alibaba just gave away a model that matches Claude's flagship, and it literally runs on a $700 used graphics card. The Qwen models crossed 3 BILLION downloads in six months. Hugging Face counted 418 million downloads for Google this year, and 227 million for Meta. Alibaba cleared more than four times both of them combined. Then today it released Qwen3.8-27B under an Apache 2.0 license. The model has 27 billion parameters, native vision, and a 262,000 token context window. Developers are running it locally on 17 gigabytes of memory, on used cards that cost a few hundred dollars. Alibaba's own benchmark table claims it beats Opus 4.6 Max on computer use by 84.3 to 72.7, on mobile use by 81.9 to 62, and on visual math by 94.6 to 65.5. Those numbers come from the vendor and nobody has independently verified them yet, so treat them as a claim. But the generation over generation jumps are harder to wave away: On DeepSWE the score went from 13.3 to 42.2. On software engineering it went from 49.3 to 79.0. That happened in ONE release cycle. And Apache 2.0 means anyone can download the weights, modify them, build products on them, sell those products, and never pay or ask permission. It cannot be revoked. Once the file is on your drive it is yours permanently. 3 billion downloads means those files already sit on machines in every country on Earth. Alibaba could delete everything tomorrow and it would change nothing. Washington spent 4 years building an export control regime around chips, model weights, and entity lists. Every piece of it assumes a chokepoint exists somewhere. A fab, a shipment, a company that can be told no. But there is no chokepoint for a file that has already been copied three billion times. And the copying compounds. Hugging Face counted 151,448 models built on top of Qwen, which is 2.6x Meta's entire footprint and 4.7x the number of Llama repositories. New ones appear at roughly 200 a day. The report says Qwen has become "part of the default workflow for developers deciding what models to fine-tune and deploy." Alibaba is also pushing Qwen through its cloud into Southeast Asia and Africa, markets where American labs have almost no presence, and where a very large share of the next generation of developers will learn to build. Meta and Nvidia have both rushed out new open models in recent weeks. That is what a response looks like when you feel the floor move. And to be clear, these are download and derivative numbers, not usage numbers. ChatGPT and Claude cannot be downloaded at all, so they do not appear in this comparison. What the figures measure is what developers choose to build on top of, which is a different question from what consumers type into a box. That is also why it matters MORE. Consumer habits change in an afternoon. Infrastructure choices last a decade, because everything built on top has to be rewritten to undo them. The American labs are valued on an assumption that frontier intelligence stays scarce, expensive, and rented by the token. Alibaba just made a version of it free, permanent, and small enough to run on hardware people already own. You will not get an announcement when the software you use every day starts running on a Chinese model underneath. Go and count how many of the tools you rely on could be rebuilt on free weights this year.

Ricardo

81,936 Aufrufe • vor 1 Monat

Tencent's Hy3 lands today with real coding and agent ability and a price that undercuts the models you'd put it against. It's a 295B MoE with only 21B active parameters, a big part of why it stays cheap to serve, plus a 256K context window and an Apache 2.0 license that even outperforms GLM-5.1 although GLM is 2.5x larger. On OpenRouter it runs at $0.14 per million input tokens and $0.58 per million output. For a model Tencent benchmarks against much larger flagships, that pricing is insane. Price-performance-ratio is China's real moat. I already wrote about Tencent Hy3 earlier and the significant leap in quality it represents. Therefore, I'm pleased to recommend this model from Tencent Hy. Open Router link down below: #Hy3 #Hunyuan #AICoding
0:45

Tencent's Hy3 lands today with real coding and agent ability and a price that undercuts the models you'd put it against. It's a 295B MoE with only 21B active parameters, a big part of why it stays cheap to serve, plus a 256K context window and an Apache 2.0 license that even outperforms GLM-5.1 although GLM is 2.5x larger. On OpenRouter it runs at $0.14 per million input tokens and $0.58 per million output. For a model Tencent benchmarks against much larger flagships, that pricing is insane. Price-performance-ratio is China's real moat. I already wrote about Tencent Hy3 earlier and the significant leap in quality it represents. Therefore, I'm pleased to recommend this model from Tencent Hy. Open Router link down below: #Hy3 #Hunyuan #AICoding

Chubby♨️

23,439 Aufrufe • vor 2 Monaten

Run your coding agent on a local model, with QVAC. OpenCode now plugs into a model on your own machine through our official plugin. That changes two things. Privacy by default. Your repo, your prompts, and every command the agent runs stay on your device. No third-party model ever sees your code. For an agent that reads your codebase and runs commands, that is the point, not a nice-to-have. New workflows. With no meter and no rate limit, the agent does not have to stop. Let it loop on a task, refactor a whole repo overnight, or run a watch-build-fix loop that keeps going. Setup takes a few minutes: one line in opencode.json and OpenCode runs on a local QVAC model. In the video demo, it handles three real developer tasks back to back, fully offline.
1:53

Run your coding agent on a local model, with QVAC. OpenCode now plugs into a model on your own machine through our official plugin. That changes two things. Privacy by default. Your repo, your prompts, and every command the agent runs stay on your device. No third-party model ever sees your code. For an agent that reads your codebase and runs commands, that is the point, not a nice-to-have. New workflows. With no meter and no rate limit, the agent does not have to stop. Let it loop on a task, refactor a whole repo overnight, or run a watch-build-fix loop that keeps going. Setup takes a few minutes: one line in opencode.json and OpenCode runs on a local QVAC model. In the video demo, it handles three real developer tasks back to back, fully offline.

QVAC

11,382,936 Aufrufe • vor 3 Monaten

The entire AI industry is racing to build the smartest model. Satya Nadella just admitted that is not where the money is. The model is not the product. The harness is. That is the exact line. And it changes what Microsoft is actually competing on. OpenAI, Anthropic, Google, xAI, Meta every frontier lab is pouring hundreds of billions into training compute, chasing the next capability jump. Each betting that raw model intelligence is the moat. Microsoft is doing the opposite. It is building the harness the orchestration layer that sits above the model, connecting it to tools, data, permissions, sub-agents, and enterprise workflows. And it is letting OpenAI, Anthropic, and MAI compete to plug into it. "You need the model. But the model is not the product. The harness is." So do the math on what a harness actually does. A raw model dropped into an enterprise answers questions. That is a chatbot. A harness turns that same model into an agent that reads the SharePoint, edits the ERP entry, pulls the GitHub PR, updates Salesforce, and files the Excel report with the right permissions, the right audit trail, and the right sub-agent for each sub-task. The model provides the intelligence. The harness converts intelligence into work. Now here's where it gets interesting. "Even the best model in the world will feel broken without a great harness. And an okay model with a great harness can feel like magic." If that is true, the enterprise buyer is not buying model quality. The enterprise buyer is buying the harness. Which means model quality becomes a commodity input over time, and harness quality becomes the sustainable moat. Compare that to the strategy the entire frontier lab industry is executing. Everyone else is chasing the numerator raw intelligence. Almost nobody at scale is racing to build the denominator the orchestration layer that determines whether that intelligence can actually be deployed profitably inside a real company. The frontier model race has a 10 to 20 percent chance of producing a single dominant winner. Nadella just told the industry he does not need to be that winner. If OpenAI wins, Microsoft wins. If Anthropic wins, Microsoft wins. If MAI wins, Microsoft wins. If someone Microsoft has never heard of trains a better model in 2027, Microsoft still wins. Because the compute they train on, the harness they get plugged into, the enterprise contracts they get delivered through, and the products they sit inside are all Microsoft. He is not building the best AI model. He is building the layer that the best AI model has to run on to make anyone money. I wonder which position looks more valuable in ten years.
3:19

The entire AI industry is racing to build the smartest model. Satya Nadella just admitted that is not where the money is. The model is not the product. The harness is. That is the exact line. And it changes what Microsoft is actually competing on. OpenAI, Anthropic, Google, xAI, Meta every frontier lab is pouring hundreds of billions into training compute, chasing the next capability jump. Each betting that raw model intelligence is the moat. Microsoft is doing the opposite. It is building the harness the orchestration layer that sits above the model, connecting it to tools, data, permissions, sub-agents, and enterprise workflows. And it is letting OpenAI, Anthropic, and MAI compete to plug into it. "You need the model. But the model is not the product. The harness is." So do the math on what a harness actually does. A raw model dropped into an enterprise answers questions. That is a chatbot. A harness turns that same model into an agent that reads the SharePoint, edits the ERP entry, pulls the GitHub PR, updates Salesforce, and files the Excel report with the right permissions, the right audit trail, and the right sub-agent for each sub-task. The model provides the intelligence. The harness converts intelligence into work. Now here's where it gets interesting. "Even the best model in the world will feel broken without a great harness. And an okay model with a great harness can feel like magic." If that is true, the enterprise buyer is not buying model quality. The enterprise buyer is buying the harness. Which means model quality becomes a commodity input over time, and harness quality becomes the sustainable moat. Compare that to the strategy the entire frontier lab industry is executing. Everyone else is chasing the numerator raw intelligence. Almost nobody at scale is racing to build the denominator the orchestration layer that determines whether that intelligence can actually be deployed profitably inside a real company. The frontier model race has a 10 to 20 percent chance of producing a single dominant winner. Nadella just told the industry he does not need to be that winner. If OpenAI wins, Microsoft wins. If Anthropic wins, Microsoft wins. If MAI wins, Microsoft wins. If someone Microsoft has never heard of trains a better model in 2027, Microsoft still wins. Because the compute they train on, the harness they get plugged into, the enterprise contracts they get delivered through, and the products they sit inside are all Microsoft. He is not building the best AI model. He is building the layer that the best AI model has to run on to make anyone money. I wonder which position looks more valuable in ten years.

Vikram M

21,463 Aufrufe • vor 2 Monaten

CONOR IS NOT SELLING THE OLD CONOR ACT THIS WEEK. Everyone waited for the insults. Instead, he is talking about Max with respect. That matters. Because the loud version of Conor is easy to understand. The disciplined version is harder to price in. If he is really living in the gym, that is not just promo. That is a fighter trying to remove the biggest question around him. Not power. Not fame. Discipline. But this is still not a fantasy comeback story. Max is active. Max is in shape. Max has been living in real fight rhythm while Conor has been rebuilding from the outside. Welterweight is new for Max, and that cuts both ways. The size may feel different. The strength may feel different. But no weight cut could also unlock a cleaner version of him. More fuel. More durability. Less fight-week damage before the fight even starts. That is why this should look competitive. Not because Conor talks. Not because nostalgia needs it. Not because the UFC needs a movie ending. Because Conor looks serious, and Max is still a dangerous active fighter. We find out on Saturday which part matters more.
4:20

CONOR IS NOT SELLING THE OLD CONOR ACT THIS WEEK. Everyone waited for the insults. Instead, he is talking about Max with respect. That matters. Because the loud version of Conor is easy to understand. The disciplined version is harder to price in. If he is really living in the gym, that is not just promo. That is a fighter trying to remove the biggest question around him. Not power. Not fame. Discipline. But this is still not a fantasy comeback story. Max is active. Max is in shape. Max has been living in real fight rhythm while Conor has been rebuilding from the outside. Welterweight is new for Max, and that cuts both ways. The size may feel different. The strength may feel different. But no weight cut could also unlock a cleaner version of him. More fuel. More durability. Less fight-week damage before the fight even starts. That is why this should look competitive. Not because Conor talks. Not because nostalgia needs it. Not because the UFC needs a movie ending. Because Conor looks serious, and Max is still a dangerous active fighter. We find out on Saturday which part matters more.

Afanasiev Ivan

148,145 Aufrufe • vor 2 Monaten

watch anon! 27 billion parameter model dancing at 42 tokens a second on a used rtx 3060ti, that costs 200 dollars. consuming 159 watts. temp 62 degrees. pushing almost 100% on the gpu. 6.8 of 8 gigs used with a 128k context window loaded. and every one of those spikes is the model reasoning and firing tool calls. the model is bonsai 27b at 1 bit. PrismML took qwen 3.6 27b, and crushed it to 3.9 gigabytes, about a single sign bit per weight. it drops onto an 8gb rtx 3060 ti, an ampere card from 2020, and it doesn't just load, it serves, it holds the full 128k window, it runs a hermes agent for half an hour straight, it hits its own failures and recovers, it never once drops the loop. the absurd part is the power meter. that same 27b, in its normal quant 4, wants a 24gb card and nearly 17 gigs of weights. this is it at 3.9, doing agentic work at 160 watts, on the tier people call budget. the numbers are sitting right there on nvtop, not in a slide, not a claim, the meter reading while it works. so i keep landing on one question. the model is open, apache 2.0. the card is old and cheap. the whole thing runs offline. why is almost nobody actually running this. is it the hardware, the setup, or did nobody tell you a 27b fits on the card already in your closet.
0:26

watch anon! 27 billion parameter model dancing at 42 tokens a second on a used rtx 3060ti, that costs 200 dollars. consuming 159 watts. temp 62 degrees. pushing almost 100% on the gpu. 6.8 of 8 gigs used with a 128k context window loaded. and every one of those spikes is the model reasoning and firing tool calls. the model is bonsai 27b at 1 bit. PrismML took qwen 3.6 27b, and crushed it to 3.9 gigabytes, about a single sign bit per weight. it drops onto an 8gb rtx 3060 ti, an ampere card from 2020, and it doesn't just load, it serves, it holds the full 128k window, it runs a hermes agent for half an hour straight, it hits its own failures and recovers, it never once drops the loop. the absurd part is the power meter. that same 27b, in its normal quant 4, wants a 24gb card and nearly 17 gigs of weights. this is it at 3.9, doing agentic work at 160 watts, on the tier people call budget. the numbers are sitting right there on nvtop, not in a slide, not a claim, the meter reading while it works. so i keep landing on one question. the model is open, apache 2.0. the card is old and cheap. the whole thing runs offline. why is almost nobody actually running this. is it the hardware, the setup, or did nobody tell you a 27b fits on the card already in your closet.

Sudo su

23,453 Aufrufe • vor 2 Monaten

TWO BOXES THE SIZE OF A MAC MINI JUST RAN A 235 BILLION PARAMETER MODEL ON A DESK It is two NVIDIA DGX Spark units linked by a single cable. A year ago a model this size meant renting a GPU cluster by the hour. Now it sits next to your monitor for around $8,000. Here is the twist most people miss. Linking them does not create one shared 256GB memory pool. The model is split across both boxes, and that is the only reason a 235B model fits at all. It answers at roughly 10 tokens per second, and both chips sit at just 74 degrees while sipping around 50 watts. Every token stays on the desk. Nothing touches a cloud, and nothing leaves the room. The ceiling for what you can run at home just jumped from 70B to 235B. Bookmark this & Watch it run ↓
1:21

TWO BOXES THE SIZE OF A MAC MINI JUST RAN A 235 BILLION PARAMETER MODEL ON A DESK It is two NVIDIA DGX Spark units linked by a single cable. A year ago a model this size meant renting a GPU cluster by the hour. Now it sits next to your monitor for around $8,000. Here is the twist most people miss. Linking them does not create one shared 256GB memory pool. The model is split across both boxes, and that is the only reason a 235B model fits at all. It answers at roughly 10 tokens per second, and both chips sit at just 74 degrees while sipping around 50 watts. Every token stays on the desk. Nothing touches a cloud, and nothing leaves the room. The ceiling for what you can run at home just jumped from 70B to 235B. Bookmark this & Watch it run ↓

slash1s

101,317 Aufrufe • vor 3 Monaten