Загрузка видео...

Не удалось загрузить видео

На главную

🎉 Congrats to MiniMax (official) on releasing MiniMax M3! Frontier coding and agentic capabilities, native image and video input, computer use, and a 1M-token context window, all in a single open model. At the heart of M3 is MSA, a new sparse attention architecture: instead of attending densely over...

vllm_project's profile picture

vLLM

43,651 subscribers

40,306 просмотров • 2 месяцев назад •via X (Twitter)

ИгрыНаука и технологии
LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

With Kimi K3 Day-0 on vLLM: Open Frontier Intelligence for Everyone 🚀 At 2.8 trillion parameters, Moonshot AI's Kimi K3 is one of the most powerful open-weight models ever released. Starting today, you can serve it on vLLM the moment the weights are public. What K3 brings: 🧠 2.8T-parameter Mixture-of-Experts (16 of 896 experts active per token) 📚 1M-token context window 🛠️ Native multimodal understanding, including vision ⚡ Kimi Delta Attention: a hybrid of linear and full attention that makes million-token context affordable Huge thank you to Kimi.ai AI for the model release and partnership, Inferact for leading the vLLM optimizations, and to our partners at NVIDIA, AMD, and the broader vLLM community. 1/6
0:30

With Kimi K3 Day-0 on vLLM: Open Frontier Intelligence for Everyone 🚀 At 2.8 trillion parameters, Moonshot AI's Kimi K3 is one of the most powerful open-weight models ever released. Starting today, you can serve it on vLLM the moment the weights are public. What K3 brings: 🧠 2.8T-parameter Mixture-of-Experts (16 of 896 experts active per token) 📚 1M-token context window 🛠️ Native multimodal understanding, including vision ⚡ Kimi Delta Attention: a hybrid of linear and full attention that makes million-token context affordable Huge thank you to Kimi.ai AI for the model release and partnership, Inferact for leading the vLLM optimizations, and to our partners at NVIDIA, AMD, and the broader vLLM community. 1/6

vLLM

102,101 просмотров • 19 дней назад

I plugged MiniMax (official) M3 into Claude Code in under 5 minutes and gave it a real task. most models need hand-holding every few steps. M3 ran 147 iterations on its own. what it actually did: → took a broken CUDA kernel → ran it for 24 hours, autonomously → pushed hardware efficiency from 7.6% to 71.3% → shipped a 9.4x speedup I didn't touch it once across that whole window. the rest of the spec sheet: → 1M token context window → native multimodal from day one → fully open weights pricing: → $20/month for 1.7B tokens → $50/month for 5.1B tokens → text, image, speech, and music all share one pool usually with this kind of capability the bill catches up fast. M3 keeps both lines low at the same time. that's the part that surprised me. and if this is what M3 does at launch, the full rollout is going to be one to watch. what would you build with 1M context? #MiniMax #MiniMaxM3 #ClaudeCode #LLM
0:25

I plugged MiniMax (official) M3 into Claude Code in under 5 minutes and gave it a real task. most models need hand-holding every few steps. M3 ran 147 iterations on its own. what it actually did: → took a broken CUDA kernel → ran it for 24 hours, autonomously → pushed hardware efficiency from 7.6% to 71.3% → shipped a 9.4x speedup I didn't touch it once across that whole window. the rest of the spec sheet: → 1M token context window → native multimodal from day one → fully open weights pricing: → $20/month for 1.7B tokens → $50/month for 5.1B tokens → text, image, speech, and music all share one pool usually with this kind of capability the bill catches up fast. M3 keeps both lines low at the same time. that's the part that surprised me. and if this is what M3 does at launch, the full rollout is going to be one to watch. what would you build with 1M context? #MiniMax #MiniMaxM3 #ClaudeCode #LLM

Nelly;

17,576 просмотров • 2 месяцев назад

MiniMax (official) H3 is live in SGLang Diffusion, with day-0 serving support 🎬 This open model matches Seedance 2.0 at 1/3 the cost, or $0 if you run it locally on 2x 5090 or 1 RTX 6000. With SGLang Diffusion, you can build visual concepts, motion design, e-commerce creatives, precise video edits, animation and stylized visuals, all locally on your own machine. No API bill, no waitlist. H3 takes text, image, video and audio in a single context, and generates 5-15s clips at native 2K, 24fps with native stereo audio. SGLang Diffusion also runs on NVIDIA AI Blackwell and Hopper, and AMD MI355X / MI300X. Let's create something with MiniMax H3 👇
0:31

MiniMax (official) H3 is live in SGLang Diffusion, with day-0 serving support 🎬 This open model matches Seedance 2.0 at 1/3 the cost, or $0 if you run it locally on 2x 5090 or 1 RTX 6000. With SGLang Diffusion, you can build visual concepts, motion design, e-commerce creatives, precise video edits, animation and stylized visuals, all locally on your own machine. No API bill, no waitlist. H3 takes text, image, video and audio in a single context, and generates 5-15s clips at native 2K, 24fps with native stereo audio. SGLang Diffusion also runs on NVIDIA AI Blackwell and Hopper, and AMD MI355X / MI300X. Let's create something with MiniMax H3 👇

LMSYS Org

96,841 просмотров • 12 дней назад

🎉 Day-0 support for in vLLM, available today in v0.23.0! Congrats to Z.ai on GLM-5.2, a flagship model built for long-horizon coding agents. ✨ 1M-token context, built to hold project-scale engineering work in a single run ✨ Tuned for long-horizon coding: large-scale implementation, automated research, and performance optimization ✨ One task can carry a full dev workflow, from requirements to a deployable product across platforms ✨ Client-side and mobile engineering, including an on-device debugging loop Try it out running it on vLLM today: 🔗
0:30

🎉 Day-0 support for in vLLM, available today in v0.23.0! Congrats to Z.ai on GLM-5.2, a flagship model built for long-horizon coding agents. ✨ 1M-token context, built to hold project-scale engineering work in a single run ✨ Tuned for long-horizon coding: large-scale implementation, automated research, and performance optimization ✨ One task can carry a full dev workflow, from requirements to a deployable product across platforms ✨ Client-side and mobile engineering, including an on-device debugging loop Try it out running it on vLLM today: 🔗

vLLM

36,240 просмотров • 2 месяцев назад

Everyone's sleeping on MiniMax. Again. They just shipped M3. The first open-weights model to combine frontier coding, 1M context, and native multimodality in one drop. I plugged it into Claude Code this morning. Pasted a design from Dribbble. Watched M3 write production-ready React code in one session. At the agency, I just replaced Opus 4.8 with M3 for 80% of our coding tasks. The output is the same and we are running everything at a fraction of the cost. Open infrastructure is the future.
1:35

Everyone's sleeping on MiniMax. Again. They just shipped M3. The first open-weights model to combine frontier coding, 1M context, and native multimodality in one drop. I plugged it into Claude Code this morning. Pasted a design from Dribbble. Watched M3 write production-ready React code in one session. At the agency, I just replaced Opus 4.8 with M3 for 80% of our coding tasks. The output is the same and we are running everything at a fraction of the cost. Open infrastructure is the future.

Prajwal Tomar

12,904 просмотров • 2 месяцев назад

Inkling, Thinking Machines' first open model, dropped today: 975B total / 41B active MoE, up to 1M context, reasoning natively over text, images, and audio. Serving and RL support are already live: you can run and shape it on an open stack, starting now. Day 0 support on SGLang SGLang and Miles RadixArk👇 - Inkling's new architecture (ShortConv, attention with relative positional embedding, shared expert sink MoE) is natively implemented and deeply optimized, with prefill full CUDA graph and MXFP8 KV cache - Full parameter and LoRA RL in a customized Megatron backend, train inference consistency via customized kernels, routing replay, and cross-runtime parameter synchronization - DFlash speculative decoding from Modal for low-latency serving Launching now, blog and cookbook in the comments ⬇️
1:01

Inkling, Thinking Machines' first open model, dropped today: 975B total / 41B active MoE, up to 1M context, reasoning natively over text, images, and audio. Serving and RL support are already live: you can run and shape it on an open stack, starting now. Day 0 support on SGLang SGLang and Miles RadixArk👇 - Inkling's new architecture (ShortConv, attention with relative positional embedding, shared expert sink MoE) is natively implemented and deeply optimized, with prefill full CUDA graph and MXFP8 KV cache - Full parameter and LoRA RL in a customized Megatron backend, train inference consistency via customized kernels, routing replay, and cross-runtime parameter synchronization - DFlash speculative decoding from Modal for low-latency serving Launching now, blog and cookbook in the comments ⬇️

LMSYS Org

145,330 просмотров • 1 месяц назад

SGLang day-0 speed on Kimi K3: 423 tok/s (measured on gsm8k), plus RL support ready in Miles RadixArk! How the largest open-source model runs this fast: we natively implemented and deeply optimized K3’s new architecture with fused KDA decode kernels, DP attention, DSpark, PD disagg, and KDA-aware prefix caching. We've passed Kimi Vendor Verifier and are ready for production! Thanks to Kimi.ai, NVIDIA, AMD, KVCache.AI, Modal, and Baseten for building this with us, and to Google Cloud, Nebius Token Factory, fal, DigitalOcean, Runpod, DeepInfra and GMI Cloud for serving K3 on SGLang. Blog, cookbook, benchmarks in the comments. P.S. This demo video? Kimi K3 made it itself. Play the game 👇
0:52

SGLang day-0 speed on Kimi K3: 423 tok/s (measured on gsm8k), plus RL support ready in Miles RadixArk! How the largest open-source model runs this fast: we natively implemented and deeply optimized K3’s new architecture with fused KDA decode kernels, DP attention, DSpark, PD disagg, and KDA-aware prefix caching. We've passed Kimi Vendor Verifier and are ready for production! Thanks to Kimi.ai, NVIDIA, AMD, KVCache.AI, Modal, and Baseten for building this with us, and to Google Cloud, Nebius Token Factory, fal, DigitalOcean, Runpod, DeepInfra and GMI Cloud for serving K3 on SGLang. Blog, cookbook, benchmarks in the comments. P.S. This demo video? Kimi K3 made it itself. Play the game 👇

LMSYS Org

201,687 просмотров • 19 дней назад

MiniMax H3 is native in ComfyUI, day zero with the weights. Model highlights: → Text-to-video, prompt only → Image-to-video → First-and-last-frame - control the opening frame, the closing frame, or both → Reference-to-video - carry a subject, a motion, or a voice through the clip → In-place editing - modify a shot you already have instead of regenerating it The capability MiniMax (official) leads with is multimodal context understanding, and it's what collapses those five tasks into one model. H3 takes images, audio, and video together and resolves them against a prompt that explains how they relate. Download the workflows & learn more in our blog 👇
1:06

MiniMax H3 is native in ComfyUI, day zero with the weights. Model highlights: → Text-to-video, prompt only → Image-to-video → First-and-last-frame - control the opening frame, the closing frame, or both → Reference-to-video - carry a subject, a motion, or a voice through the clip → In-place editing - modify a shot you already have instead of regenerating it The capability MiniMax (official) leads with is multimodal context understanding, and it's what collapses those five tasks into one model. H3 takes images, audio, and video together and resolves them against a prompt that explains how they relate. Download the workflows & learn more in our blog 👇

ComfyUI

229,553 просмотров • 12 дней назад

A tricky LLM interview question: You're serving a reasoning model on vLLM, and it keeps running out of GPU memory on long traces. So you add KV cache compression and evict 90% of the cached tokens. VRAM usage stays as is and GPU still runs out of memory. Why? (answer below) Evicting 90% of the KV cache can free almost none of the memory it was using. This sounds counterintuitive, but it follows directly from how production servers store the cache today. The KV cache grows with every token a model generates. Each token appends its key and value vectors across every layer, and nothing is freed while generation continues. This is the dominant memory cost for reasoning models. If a 32K-token CoT caches ~32K tokens of KV vectors, a Qwen3-32B with 4-bit weights will run out-of-memory around 24K tokens on a 24GB GPU. One obvious solution is to keep the important tokens and drop the rest, since attention is sparse enough to allow it. But this does not solve the memory problem yet. The reason is paged attention, which is the memory manager behind vLLM and most production servers. Under the hood, it splits GPU memory into fixed physical blocks, each one holds the KV for about 16 tokens. This block returns to the allocator only when every slot inside it is empty. Since the eviction logic selects tokens by importance, and such tokens are scattered across blocks... ...so despite eviction, almost every block is left with at least some survivor tokens. For instance, if the logic evicts 14k of 16k tokens across 1,000 blocks, most likely every block will still have a token. This means the allocator frees almost nothing. Placing the new tokens into those freed slots is not ideal because it breaks the cache's layout. Say token 16,001 arrives, and it's placed in the slot the 40th token used to hold. The cache now reads position 38, then 16,001, then 41, so the cache is no longer in token order. Attention can still compute the right answer from that, but only if every slot now carries a separate note recording which position it actually holds. This introduces another bookkeeping cost that an in-order layout inherently avoids. So the cache is logically 90% smaller and still physically the same size. Many compression results miss this because they measure on pre-allocated contiguous tensors rather than a paged server. There's another problem. Eviction methods pick which tokens to keep by looking at the attention scores themselves (as expected). But fast attention kernels used in production, like FlashAttention, never save those scores. They compute attention in small pieces and throw the full score grid away as they go, which is also why they're fast. So the exact signal eviction methods need isn't available in memory. The workaround is to fall back to eager attention and build the full matrix, which gives up the speed FlashAttention was there to provide. NVIDIA published a method called TriAttention to solve both these problems. It never needs attention scores. Instead, it scores tokens from the geometry of the model's key and query vectors before RoPE is applied, where those vectors sit in stable clusters. For the memory problem, it runs a compaction pass every 128 decoded tokens. The surviving tokens slide forward to close the holes eviction creates, so whole blocks empty out and return to the allocator while the cache stays in token order. On long reasoning traces, the approach matches full-attention accuracy while decoding 2.5x faster and using 10.7x less KV memory. KV cache compression is a big infrastructure problem. The number that decides whether it works is the count of freed blocks, not the count of evicted tokens. You can find the NVIDIA write-up here: I wrote a first-principles breakdown of how the KV cache works. It walks through why the model stores keys and values at all, why the cache grows with every token, and a comparison of LLM generation speed with and without KV caching. Read it below.
0:23

A tricky LLM interview question: You're serving a reasoning model on vLLM, and it keeps running out of GPU memory on long traces. So you add KV cache compression and evict 90% of the cached tokens. VRAM usage stays as is and GPU still runs out of memory. Why? (answer below) Evicting 90% of the KV cache can free almost none of the memory it was using. This sounds counterintuitive, but it follows directly from how production servers store the cache today. The KV cache grows with every token a model generates. Each token appends its key and value vectors across every layer, and nothing is freed while generation continues. This is the dominant memory cost for reasoning models. If a 32K-token CoT caches ~32K tokens of KV vectors, a Qwen3-32B with 4-bit weights will run out-of-memory around 24K tokens on a 24GB GPU. One obvious solution is to keep the important tokens and drop the rest, since attention is sparse enough to allow it. But this does not solve the memory problem yet. The reason is paged attention, which is the memory manager behind vLLM and most production servers. Under the hood, it splits GPU memory into fixed physical blocks, each one holds the KV for about 16 tokens. This block returns to the allocator only when every slot inside it is empty. Since the eviction logic selects tokens by importance, and such tokens are scattered across blocks... ...so despite eviction, almost every block is left with at least some survivor tokens. For instance, if the logic evicts 14k of 16k tokens across 1,000 blocks, most likely every block will still have a token. This means the allocator frees almost nothing. Placing the new tokens into those freed slots is not ideal because it breaks the cache's layout. Say token 16,001 arrives, and it's placed in the slot the 40th token used to hold. The cache now reads position 38, then 16,001, then 41, so the cache is no longer in token order. Attention can still compute the right answer from that, but only if every slot now carries a separate note recording which position it actually holds. This introduces another bookkeeping cost that an in-order layout inherently avoids. So the cache is logically 90% smaller and still physically the same size. Many compression results miss this because they measure on pre-allocated contiguous tensors rather than a paged server. There's another problem. Eviction methods pick which tokens to keep by looking at the attention scores themselves (as expected). But fast attention kernels used in production, like FlashAttention, never save those scores. They compute attention in small pieces and throw the full score grid away as they go, which is also why they're fast. So the exact signal eviction methods need isn't available in memory. The workaround is to fall back to eager attention and build the full matrix, which gives up the speed FlashAttention was there to provide. NVIDIA published a method called TriAttention to solve both these problems. It never needs attention scores. Instead, it scores tokens from the geometry of the model's key and query vectors before RoPE is applied, where those vectors sit in stable clusters. For the memory problem, it runs a compaction pass every 128 decoded tokens. The surviving tokens slide forward to close the holes eviction creates, so whole blocks empty out and return to the allocator while the cache stays in token order. On long reasoning traces, the approach matches full-attention accuracy while decoding 2.5x faster and using 10.7x less KV memory. KV cache compression is a big infrastructure problem. The number that decides whether it works is the count of freed blocks, not the count of evicted tokens. You can find the NVIDIA write-up here: I wrote a first-principles breakdown of how the KV cache works. It walks through why the model stores keys and values at all, why the cache grows with every token, and a comparison of LLM generation speed with and without KV caching. Read it below.

Avi Chawla

270,496 просмотров • 1 месяц назад

The MiniMax M2 model is mind-blowing! It's open-source. It outperforms Gemini 2.5, Claude 4.1, and Qwen3 across coding and tool-use benchmarks. Right now, it's one of the world's top 5 models in intelligence! And here is the best part: Claude is one of the best models you can use today, and MiniMax M2 costs only 8% of that! It's smaller, faster, and cheaper. Extremely efficient at using tokens. Minimax M2's biggest strength: High agentic capabilities. The model can plan and execute complex multi-tool workflows. It's reliable and very robust at executing long-horizon tool chains. In summary: • Low latency • Very cheap • Excels at agentic tasks • Open-source The model currently powers the MiniMax Agent and is available for a free global trial. You can access MiniMax M2's API here: To access the agent: And here is the MiniMax website: Thanks to the MiniMax team for showing me the ropes and partnering with me on this post.
6:01

The MiniMax M2 model is mind-blowing! It's open-source. It outperforms Gemini 2.5, Claude 4.1, and Qwen3 across coding and tool-use benchmarks. Right now, it's one of the world's top 5 models in intelligence! And here is the best part: Claude is one of the best models you can use today, and MiniMax M2 costs only 8% of that! It's smaller, faster, and cheaper. Extremely efficient at using tokens. Minimax M2's biggest strength: High agentic capabilities. The model can plan and execute complex multi-tool workflows. It's reliable and very robust at executing long-horizon tool chains. In summary: • Low latency • Very cheap • Excels at agentic tasks • Open-source The model currently powers the MiniMax Agent and is available for a free global trial. You can access MiniMax M2's API here: To access the agent: And here is the MiniMax website: Thanks to the MiniMax team for showing me the ropes and partnering with me on this post.

Santiago

91,197 просмотров • 9 месяцев назад

Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. Subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.
1:24

Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. Subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.

Alexander Whedon

13,274,625 просмотров • 3 месяцев назад

CHINA JUST DROPPED AN AI CODING MODEL WITH A 1M CONTEXT WINDOW. And I connected it to Claude Code to see what it could actually do. Meet GLM-X Preview On paper, a few things immediately stood out: → 1M context window → Agentic coding capabilities → Works inside Claude Code → Designed for large-scale coding and reasoning workflows But specs don't matter much if the model can't deliver in practice. So I gave it a real-world task. THE TEST One prompt: > Build a modern AI lead generation dashboard using React and Tailwind CSS. Requirements: → Dark mode → Analytics dashboard → Lead table → Email outreach section → Responsive design → Production-ready component structure Instead of generating a few snippets, it planned the architecture, generated the dashboard components, created the Tailwind configuration, and walked through the implementation requirements. What impressed me most wasn't the code itself. It was how well it maintained context throughout the workflow. That's where a 1M context window starts becoming useful. Less time re-explaining requirements. Less context loss. More room for complex projects. The AI coding race is getting very interesting. And it's no longer just GPT, Claude, and Gemini competing for attention. Results from my test below 👇
0:33

CHINA JUST DROPPED AN AI CODING MODEL WITH A 1M CONTEXT WINDOW. And I connected it to Claude Code to see what it could actually do. Meet GLM-X Preview On paper, a few things immediately stood out: → 1M context window → Agentic coding capabilities → Works inside Claude Code → Designed for large-scale coding and reasoning workflows But specs don't matter much if the model can't deliver in practice. So I gave it a real-world task. THE TEST One prompt: > Build a modern AI lead generation dashboard using React and Tailwind CSS. Requirements: → Dark mode → Analytics dashboard → Lead table → Email outreach section → Responsive design → Production-ready component structure Instead of generating a few snippets, it planned the architecture, generated the dashboard components, created the Tailwind configuration, and walked through the implementation requirements. What impressed me most wasn't the code itself. It was how well it maintained context throughout the workflow. That's where a 1M context window starts becoming useful. Less time re-explaining requirements. Less context loss. More room for complex projects. The AI coding race is getting very interesting. And it's no longer just GPT, Claude, and Gemini competing for attention. Results from my test below 👇

Md Riyazuddin

31,199 просмотров • 1 месяц назад

Serving GLM5.2 NVFP4 Agentic Workload with SGLang: How We Reached 500 TPS on 8xB300 at bs=1 In this deep dive, we explain how SGLang reaches 500+ tok/s/user at bs=1 on 8xB300, with 18 to 34% higher single-user interactivity within two weeks since day-0, and 6 to 11% better peak throughput at high concurrency, benchmarked on a real multi-turn agentic coding workload. Our new TopK-V2 kernel is 2.33x faster at 80K ISL, scaling to 10.17x at 1M ISL, keeping interactivity essentially flat out to 1M tokens. Part of the story is the architecture itself. GLM-5.2 applies IndexShare to its DSA layers and ships a stronger MTP head reusing IndexShare and KVShare. The rest comes from our serving optimizations. Special thanks to NVIDIA AI for the help in day-0 support of GLM-5.2 NVFP4, and to Z.ai for IndexShare in SGLang!
0:47

Serving GLM5.2 NVFP4 Agentic Workload with SGLang: How We Reached 500 TPS on 8xB300 at bs=1 In this deep dive, we explain how SGLang reaches 500+ tok/s/user at bs=1 on 8xB300, with 18 to 34% higher single-user interactivity within two weeks since day-0, and 6 to 11% better peak throughput at high concurrency, benchmarked on a real multi-turn agentic coding workload. Our new TopK-V2 kernel is 2.33x faster at 80K ISL, scaling to 10.17x at 1M ISL, keeping interactivity essentially flat out to 1M tokens. Part of the story is the architecture itself. GLM-5.2 applies IndexShare to its DSA layers and ships a stronger MTP head reusing IndexShare and KVShare. The rest comes from our serving optimizations. Special thanks to NVIDIA AI for the help in day-0 support of GLM-5.2 NVFP4, and to Z.ai for IndexShare in SGLang!

LMSYS Org

106,886 просмотров • 1 месяц назад

🚨 ALIBABA JUST OPEN-SOURCED THE MOST POWERFUL MODEL IN THE WORLD - Qwen3.5 plus And it changes the AI race forever. This is not another “bigger model” release. This is a complete architectural reset. 397B parameters → only 17B active at inference, Qwen3.5 plus delivers performance on par with GPT-5.2 and Gemini-3-Pro. A new Hybrid Sparse MoE + Linear Attention architecture + Multi-Token Prediction = near-instant long-context responses and massive speed gains. Built as a true native multimodal model, it doesn’t just chat it builds. With a single prompt: “Use the React framework to create a visually impressive Mapbox demo featuring locations in Beijing and Shanghai” it generates a live interactive 3D map UI, structured components, real layouts, and production-ready frontend logic. This is real agentic, multimodal coding not a static code block. MMLU-Pro: 87.8 GPQA: 88.4 IFBench: 76.5 ⚡ up to 8× faster inference And it’s fully open-sourced under Apache-2.0. No locked APIs. No restricted tiers. 201 languages supported bringing frontier AI to developers, startups, and researchers worldwide. The race is no longer about the biggest model. It’s about the most usable intelligence. Open source just entered the top tier.
1:50

🚨 ALIBABA JUST OPEN-SOURCED THE MOST POWERFUL MODEL IN THE WORLD - Qwen3.5 plus And it changes the AI race forever. This is not another “bigger model” release. This is a complete architectural reset. 397B parameters → only 17B active at inference, Qwen3.5 plus delivers performance on par with GPT-5.2 and Gemini-3-Pro. A new Hybrid Sparse MoE + Linear Attention architecture + Multi-Token Prediction = near-instant long-context responses and massive speed gains. Built as a true native multimodal model, it doesn’t just chat it builds. With a single prompt: “Use the React framework to create a visually impressive Mapbox demo featuring locations in Beijing and Shanghai” it generates a live interactive 3D map UI, structured components, real layouts, and production-ready frontend logic. This is real agentic, multimodal coding not a static code block. MMLU-Pro: 87.8 GPQA: 88.4 IFBench: 76.5 ⚡ up to 8× faster inference And it’s fully open-sourced under Apache-2.0. No locked APIs. No restricted tiers. 201 languages supported bringing frontier AI to developers, startups, and researchers worldwide. The race is no longer about the biggest model. It’s about the most usable intelligence. Open source just entered the top tier.

RAVI KUMAR SAHU

380,123 просмотров • 6 месяцев назад

New short course: LLMs as Operating Systems: Agent Memory, created with Letta, and taught by its founders Charles Packer and Sarah Wooders. An LLM's input context window has limited space. Using a longer input context also costs more and results in slower processing. So, managing what's stored in this context window is important. In the innovative paper MemGPT: Towards LLMs as Operating Systems, its authors (which include the instructors) proposed using an LLM agent to manage this context window. Their system uses a large persistent memory that stores everything that could be included in the input context, and an agent decides what is actually included. Take the example of building a chatbot that needs to remember what's been said earlier in a conversation (perhaps over many days of interaction with a user). As the conversation's length grows, the memory management agent will move information from the input context to a persistent searchable database; summarize information to keep relevant facts in the input context; and restore relevant conversation elements from further back in time. This allows a chatbot to keep what's currently most relevant in its input context memory to generate the next response. When I read the original MemGPT paper, I thought it was an innovative technique for handling memory for LLMs. The open-source Letta framework, which we'll use in this course, makes MemGPT easy to implement. It adds memory to your LLM agents and gives them transparent long-term memory. In detail, you’ll learn: - How to build an agent that can edit its own limited input context memory, using tools and multi-step reasoning - What is a memory hierarchy (an idea from computer operating systems, which use a cache to speed up memory access), and how these ideas apply to managing the LLM input context (where the input context window is a "cache" storing the most relevant information; and an agent decides what to move in and out of this to/from a larger persistent storage system) - How to implement multi-agent collaboration by letting different agents share blocks of memory This course will give you a sophisticated understanding of memory management for LLMs, which is important for chatbots having long conversations, and for complex agentic workflows. Please sign up here!
2:57

New short course: LLMs as Operating Systems: Agent Memory, created with Letta, and taught by its founders Charles Packer and Sarah Wooders. An LLM's input context window has limited space. Using a longer input context also costs more and results in slower processing. So, managing what's stored in this context window is important. In the innovative paper MemGPT: Towards LLMs as Operating Systems, its authors (which include the instructors) proposed using an LLM agent to manage this context window. Their system uses a large persistent memory that stores everything that could be included in the input context, and an agent decides what is actually included. Take the example of building a chatbot that needs to remember what's been said earlier in a conversation (perhaps over many days of interaction with a user). As the conversation's length grows, the memory management agent will move information from the input context to a persistent searchable database; summarize information to keep relevant facts in the input context; and restore relevant conversation elements from further back in time. This allows a chatbot to keep what's currently most relevant in its input context memory to generate the next response. When I read the original MemGPT paper, I thought it was an innovative technique for handling memory for LLMs. The open-source Letta framework, which we'll use in this course, makes MemGPT easy to implement. It adds memory to your LLM agents and gives them transparent long-term memory. In detail, you’ll learn: - How to build an agent that can edit its own limited input context memory, using tools and multi-step reasoning - What is a memory hierarchy (an idea from computer operating systems, which use a cache to speed up memory access), and how these ideas apply to managing the LLM input context (where the input context window is a "cache" storing the most relevant information; and an agent decides what to move in and out of this to/from a larger persistent storage system) - How to implement multi-agent collaboration by letting different agents share blocks of memory This course will give you a sophisticated understanding of memory management for LLMs, which is important for chatbots having long conversations, and for complex agentic workflows. Please sign up here!

Andrew Ng

200,950 просмотров • 1 год назад

New course on serving LLMs efficiently -- how do you serve models to many concurrent users at low latency and reasonable cost? This short course is built with Red Hat and taught by Cedric Clyburn. Efficient LLM serving requires efficient memory management. A 70B-parameter model takes ~140 GB just to load the weights. On top of that, every active request needs its own chunk of GPU memory, the KV cache, to store the token context it has built up so far. In this course, you'll learn to reduce a model's memory footprint with quantization and serve it using vLLM, which handles many concurrent requests efficiently through smart memory management. Skills you'll gain: - Quantize a model and measure the accuracy tradeoff - Serve a model with vLLM and watch it handle concurrent requests efficiently - Benchmark your deployment and make informed tradeoffs between speed, cost, and accuracy Join and learn to serve LLMs efficiently:
2:11

New course on serving LLMs efficiently -- how do you serve models to many concurrent users at low latency and reasonable cost? This short course is built with Red Hat and taught by Cedric Clyburn. Efficient LLM serving requires efficient memory management. A 70B-parameter model takes ~140 GB just to load the weights. On top of that, every active request needs its own chunk of GPU memory, the KV cache, to store the token context it has built up so far. In this course, you'll learn to reduce a model's memory footprint with quantization and serve it using vLLM, which handles many concurrent requests efficiently through smart memory management. Skills you'll gain: - Quantize a model and measure the accuracy tradeoff - Serve a model with vLLM and watch it handle concurrent requests efficiently - Benchmark your deployment and make informed tradeoffs between speed, cost, and accuracy Join and learn to serve LLMs efficiently:

Andrew Ng

131,069 просмотров • 2 месяцев назад

The phone threshold is even harder than the storage number suggests. A phone exposes only part of its memory to an application, and the model must share that budget with its KV cache, activations, runtime, and the rest of the product. At 3.9 GB, 1-bit Bonsai 27B clears that threshold with room to work, making 27B-class local AI possible on a phone for the first time. (Demo Mode: cached & prefilled image context)
1:19

The phone threshold is even harder than the storage number suggests. A phone exposes only part of its memory to an application, and the model must share that budget with its KV cache, activations, runtime, and the rest of the product. At 3.9 GB, 1-bit Bonsai 27B clears that threshold with room to work, making 27B-class local AI possible on a phone for the first time. (Demo Mode: cached & prefilled image context)

PrismML

32,858 просмотров • 1 месяц назад

When will MiniMax have a Fable level model? Olive Song, research lead at MiniMax (official), joins The Bench to answer that question, plus M3, open source models closing the gap on Vibe Code Bench, and the truth about 996 culture. Full episode out now! 7:19 – Inside MiniMax's research culture: AI agents tracking papers since early last year 18:01- The headline of M3: native multimodality and one million tokens of context 21:10- The bottleneck of building models is people, not compute 22:48 – Does MiniMax work 996? 25:47 - Olive's prediction and the 30 point gap on Vibe Code Bench 32:02- Model capabilities Olive is most excited about 32:35- The lab Olive most respects outside of MiniMax 36:24- Will more labs go open source a year from now?
38:45

When will MiniMax have a Fable level model? Olive Song, research lead at MiniMax (official), joins The Bench to answer that question, plus M3, open source models closing the gap on Vibe Code Bench, and the truth about 996 culture. Full episode out now! 7:19 – Inside MiniMax's research culture: AI agents tracking papers since early last year 18:01- The headline of M3: native multimodality and one million tokens of context 21:10- The bottleneck of building models is people, not compute 22:48 – Does MiniMax work 996? 25:47 - Olive's prediction and the 30 point gap on Vibe Code Bench 32:02- Model capabilities Olive is most excited about 32:35- The lab Olive most respects outside of MiniMax 36:24- Will more labs go open source a year from now?

Vals AI

48,887 просмотров • 29 дней назад

Introducing "Building with Llama 4." This short course is created with Meta AI at Meta, and taught by Amit Sangani, Director of Partner Engineering for Meta’s AI team. Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its family of open-weight models, making them more efficient to serve. In this course, you’ll work with two of the three new models introduced in Llama 4. First is Maverick, a 400B parameter model, with 128 experts and 17B active parameters. Second is Scout, a 109B parameter model with 16 experts and 17B active parameters. Maverick and Scout support long context windows of up to a million tokens and 10M tokens, respectively. The latter is enough to support directly inputting even fairly large GitHub repos for analysis! In hands-on lessons, you’ll build apps using Llama 4’s new multimodal capabilities including reasoning across multiple images and image grounding, in which you can identify elements in images. You’ll also use the official Llama API, work with Llama 4’s long-context abilities, and learn about Llama’s newest open-source tools: its prompt optimization tool that automatically improves system prompts and synthetic data kit that generates high-quality datasets for fine-tuning. If you need an open model, Llama is a great option, and the Llama 4 family is an important part of any GenAI developer's toolkit. Through this course, you’ll learn to call Llama 4 via API, use its optimization tools, and build features that span text, images, and large context. Please sign up here:
3:22

Introducing "Building with Llama 4." This short course is created with Meta AI at Meta, and taught by Amit Sangani, Director of Partner Engineering for Meta’s AI team. Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its family of open-weight models, making them more efficient to serve. In this course, you’ll work with two of the three new models introduced in Llama 4. First is Maverick, a 400B parameter model, with 128 experts and 17B active parameters. Second is Scout, a 109B parameter model with 16 experts and 17B active parameters. Maverick and Scout support long context windows of up to a million tokens and 10M tokens, respectively. The latter is enough to support directly inputting even fairly large GitHub repos for analysis! In hands-on lessons, you’ll build apps using Llama 4’s new multimodal capabilities including reasoning across multiple images and image grounding, in which you can identify elements in images. You’ll also use the official Llama API, work with Llama 4’s long-context abilities, and learn about Llama’s newest open-source tools: its prompt optimization tool that automatically improves system prompts and synthetic data kit that generates high-quality datasets for fine-tuning. If you need an open model, Llama is a great option, and the Llama 4 family is an important part of any GenAI developer's toolkit. Through this course, you’ll learn to call Llama 4 via API, use its optimization tools, and build features that span text, images, and large context. Please sign up here:

Andrew Ng

67,846 просмотров • 1 год назад

Day 5 of why you need to use Jupiter Mobile: TRUSTED TOKEN DATA & INSIGHTS See trusted, verified token data + a clean feed of relevant content in Jupiter Mobile Token Pro’s ‘About’ section – powered by VRFD, Jupiter’s verified data layer for token information and insights. Onchain finance has never had a reliable source of truth for token information. Most wallets surface raw links or leave you to dig through socials and explorers. Instead, you get: - Simple summaries explaining what the token is and what’s currently happening - A clean feed of verified community insights, not random noise - Verified token metadata with direct links to official websites and socials If a token is covered by VRFD, you see community-validated information. If it isn’t, the full project X feed is surfaced automatically, so you still have full context. What used to take digging across multiple pages now lives directly on the token page. Just Use Jupiter (Mobile).
0:59

Day 5 of why you need to use Jupiter Mobile: TRUSTED TOKEN DATA & INSIGHTS See trusted, verified token data + a clean feed of relevant content in Jupiter Mobile Token Pro’s ‘About’ section – powered by VRFD, Jupiter’s verified data layer for token information and insights. Onchain finance has never had a reliable source of truth for token information. Most wallets surface raw links or leave you to dig through socials and explorers. Instead, you get: - Simple summaries explaining what the token is and what’s currently happening - A clean feed of verified community insights, not random noise - Verified token metadata with direct links to official websites and socials If a token is covered by VRFD, you see community-validated information. If it isn’t, the full project X feed is surfaced automatically, so you still have full context. What used to take digging across multiple pages now lives directly on the token page. Just Use Jupiter (Mobile).

Jupiter

239,841 просмотров • 7 месяцев назад