Loading video...

Video Failed to Load

Go Home

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...

lmsysorg's profile picture

LMSYS Org

17,212 subscribers

146,155 views • 1 month ago •via X (Twitter)

Science & Technology
LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

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

204,163 views • 1 month ago

🎉 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 the full KV cache, each query scores 128-token KV blocks and runs attention only over the top blocks. That is what makes 1M-token context practical to serve. M3 runs in vLLM with day-0 support, verified on NVIDIA and AMD hardware: ✨ MSA sparse attention with dedicated prefill and decode kernels ✨ 1M-token context serving with prefix caching and chunked prefill ✨ BF16 and MXFP8 checkpoints, with MoE backends for both Hopper and Blackwell ✨ Native multimodal input (image + video) ✨ Tool calling, reasoning parsing, and thinking-mode control for agent workloads Day-0 support like this is a true team effort. Grateful to the teams at MiniMax (official), NVIDIA AI, AI at AMD, and Inferact, and to the vLLM community for making it happen. 🙏 Deep dive into the implementation, kernel work, and deployment recipes: 🔗
0:21

🎉 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 the full KV cache, each query scores 128-token KV blocks and runs attention only over the top blocks. That is what makes 1M-token context practical to serve. M3 runs in vLLM with day-0 support, verified on NVIDIA and AMD hardware: ✨ MSA sparse attention with dedicated prefill and decode kernels ✨ 1M-token context serving with prefix caching and chunked prefill ✨ BF16 and MXFP8 checkpoints, with MoE backends for both Hopper and Blackwell ✨ Native multimodal input (image + video) ✨ Tool calling, reasoning parsing, and thinking-mode control for agent workloads Day-0 support like this is a true team effort. Grateful to the teams at MiniMax (official), NVIDIA AI, AI at AMD, and Inferact, and to the vLLM community for making it happen. 🙏 Deep dive into the implementation, kernel work, and deployment recipes: 🔗

vLLM

40,306 views • 2 months ago

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

99,195 views • 1 month ago

SGLang now supports DSpark, enabling confidence-driven, variable-length verification for speculative decoding 🎉 DSpark addresses a key bottleneck under load: instead of verifying every draft token, it verifies only where the draft model is confident, so the gains hold even as batch size scales. We heavily optimized variable-length verification in SGLang. Across batch sizes 1 to 256, DSpark gives the best throughput/latency tradeoff on DeepSeek-V4-Flash, ahead of both MTP and non-spec. At high concurrency, dynamic scheduling provides up to ~20% higher throughput compared to a fixed budget, while maintaining high verification quality across workloads. With fused kernels and zero-overhead scheduling, DeepSeek-V4-Pro reaches 383.7 tok/s at B=1 on B300. DSpark is now available in SGLang with support for Qwen3 and DeepSeek-V4. Thanks DeepSeek for open-sourcing! Blog with full technical details and commands to run below 👇
0:33

SGLang now supports DSpark, enabling confidence-driven, variable-length verification for speculative decoding 🎉 DSpark addresses a key bottleneck under load: instead of verifying every draft token, it verifies only where the draft model is confident, so the gains hold even as batch size scales. We heavily optimized variable-length verification in SGLang. Across batch sizes 1 to 256, DSpark gives the best throughput/latency tradeoff on DeepSeek-V4-Flash, ahead of both MTP and non-spec. At high concurrency, dynamic scheduling provides up to ~20% higher throughput compared to a fixed budget, while maintaining high verification quality across workloads. With fused kernels and zero-overhead scheduling, DeepSeek-V4-Pro reaches 383.7 tok/s at B=1 on B300. DSpark is now available in SGLang with support for Qwen3 and DeepSeek-V4. Thanks DeepSeek for open-sourcing! Blog with full technical details and commands to run below 👇

LMSYS Org

170,203 views • 1 month ago

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

68,034 views • 1 year ago

QVAC SDK 0.14.0 is live. This release makes the on-device stack faster on mobile, ships the developer-agent path, and takes local text-to-speech to 31 languages. Main highlights: - OpenCode and OpenClaw. The first official OpenCode plugin, plus a maintained OpenClaw compatibility path, both built on managed mode and qvac serve. Point a coding agent at a local model with far less setup and far fewer surprises. - Brain-computer interface transcription, on the SDK. Take recorded neural signal data and decode it into text, fully on-device, no cloud. Stream it in chunks through a simple API. In 0.14 it runs GPU-accelerated on iOS. - Text to Speech in 31 languages with our Supertonic3 upgrade. VOICE AND SPEECH - Supertonic3 multilingual TTS, 5 languages to 31. - Chatterbox and Supertonic now run on the Android GPU, with lower memory use (especially on iOS), quantized s3gen Chatterbox support, and a fix for Chatterbox occasionally emitting random speech. - Whisper transcription now runs on the iOS GPU. Parakeet runs on the Android GPU, with steadier real-time streaming. VISION AND OCR - VLM multi-tile batching: high-resolution Pan and Scan images are encoded in one pass instead of tile by tile, for faster vision throughput. - OCR on ggml (EasyOCR and DocTR) reaches full speed parity with the onnx path, across Metal, OpenCL, and Vulkan. PLATFORM AND RELIABILITY - Dynamic compute backends on Linux: one build picks the right backend at runtime, and opens the door to ROCm and CUDA support without per-backend builds. - Thinking tokens are kept out of the model context, so reasoning no longer fills the KV cache. SDK 0.14.0 is now leaner and faster to start. Let’s build.
1:51

QVAC SDK 0.14.0 is live. This release makes the on-device stack faster on mobile, ships the developer-agent path, and takes local text-to-speech to 31 languages. Main highlights: - OpenCode and OpenClaw. The first official OpenCode plugin, plus a maintained OpenClaw compatibility path, both built on managed mode and qvac serve. Point a coding agent at a local model with far less setup and far fewer surprises. - Brain-computer interface transcription, on the SDK. Take recorded neural signal data and decode it into text, fully on-device, no cloud. Stream it in chunks through a simple API. In 0.14 it runs GPU-accelerated on iOS. - Text to Speech in 31 languages with our Supertonic3 upgrade. VOICE AND SPEECH - Supertonic3 multilingual TTS, 5 languages to 31. - Chatterbox and Supertonic now run on the Android GPU, with lower memory use (especially on iOS), quantized s3gen Chatterbox support, and a fix for Chatterbox occasionally emitting random speech. - Whisper transcription now runs on the iOS GPU. Parakeet runs on the Android GPU, with steadier real-time streaming. VISION AND OCR - VLM multi-tile batching: high-resolution Pan and Scan images are encoded in one pass instead of tile by tile, for faster vision throughput. - OCR on ggml (EasyOCR and DocTR) reaches full speed parity with the onnx path, across Metal, OpenCL, and Vulkan. PLATFORM AND RELIABILITY - Dynamic compute backends on Linux: one build picks the right backend at runtime, and opens the door to ROCm and CUDA support without per-backend builds. - Thinking tokens are kept out of the model context, so reasoning no longer fills the KV cache. SDK 0.14.0 is now leaner and faster to start. Let’s build.

QVAC

23,973,950 views • 2 months ago

MolmoAct2 runs zero-shot on an SO-ARM101, no training required 🦾 Ai2 open Action Reasoning Model is now in LeRobot with the full lifecycle covered: fine-tuning (full or LoRA), evaluation, and real-robot deployment. It pairs a Molmo2-ER vision-language backbone with a flow-matching action expert to turn images, language, and proprioceptive state into action chunks. The best part: ready-made checkpoints ship with calibration correction baked in, so you can point it at your SO-100/101 and just watch it work. Inference fits in ~12GB at bf16, and LoRA fine-tuning runs on a single 24GB GPU. Big thanks to the Ai2 team for building this in the open.
0:30

MolmoAct2 runs zero-shot on an SO-ARM101, no training required 🦾 Ai2 open Action Reasoning Model is now in LeRobot with the full lifecycle covered: fine-tuning (full or LoRA), evaluation, and real-robot deployment. It pairs a Molmo2-ER vision-language backbone with a flow-matching action expert to turn images, language, and proprioceptive state into action chunks. The best part: ready-made checkpoints ship with calibration correction baked in, so you can point it at your SO-100/101 and just watch it work. Inference fits in ~12GB at bf16, and LoRA fine-tuning runs on a single 24GB GPU. Big thanks to the Ai2 team for building this in the open.

LeRobot

22,200 views • 1 month ago

Timelapse #156 (36 hrs) - Worked with the tiny corp on getting GLM 5.2 running on 8xMI300X (sglang won here) - Launched KernelBench-Mega and updated Kernelbench-Hard with h100 and b200 sweeps - Took care of boring business stuff - Did some training sweeps for specialized technical vocab audio model - Some bugs with putting kernelbench-mega and hard on cloud instances so had to do some reruns. Learned a lot though - Setting up my own local rl infra and profiling concurrency 128 rollouts with vllm. Became clear to me that I need to serve in nvfp4, use MoE only for throughput, reap so training doesn’t OOM, dig into the vllm kernel graph itself to not underutilize my hardware from poor flashinfer/cutlass selections for my rtx pro 6000 sm120 architecture - Might do online distillation from glm 5.2 but for now taking it one step at a time - Slept for a bit then woke up and showered - Fixed an issue with SGLang tensor parallel deadlock on GLM 5.2 architecture with MTP enabled - GLM 5.2 inference is 2-3x faster than coding plans and running on amd boxes - Spent time with family - Hung out with some friends - Recorded some yoctogpt lectures with the revamped notebook (high taste btw) - Setting up dflash training for GLM 5.2
0:30

Timelapse #156 (36 hrs) - Worked with the tiny corp on getting GLM 5.2 running on 8xMI300X (sglang won here) - Launched KernelBench-Mega and updated Kernelbench-Hard with h100 and b200 sweeps - Took care of boring business stuff - Did some training sweeps for specialized technical vocab audio model - Some bugs with putting kernelbench-mega and hard on cloud instances so had to do some reruns. Learned a lot though - Setting up my own local rl infra and profiling concurrency 128 rollouts with vllm. Became clear to me that I need to serve in nvfp4, use MoE only for throughput, reap so training doesn’t OOM, dig into the vllm kernel graph itself to not underutilize my hardware from poor flashinfer/cutlass selections for my rtx pro 6000 sm120 architecture - Might do online distillation from glm 5.2 but for now taking it one step at a time - Slept for a bit then woke up and showered - Fixed an issue with SGLang tensor parallel deadlock on GLM 5.2 architecture with MTP enabled - GLM 5.2 inference is 2-3x faster than coding plans and running on amd boxes - Spent time with family - Hung out with some friends - Recorded some yoctogpt lectures with the revamped notebook (high taste btw) - Setting up dflash training for GLM 5.2

Elliot Arledge

45,447 views • 2 months ago

Check out mistral.​rs, our #Rust-based open source inference engine allowing for fast #LLM serving for a variety of architectures including X-LoRA mixture-of-expert (MoE) models, Llama-3, Mistral/Mixtral, Gemma & many others. Built on the Hugging Face #Candle framework for #Rust w/ custom CUDA kernels in the backend (as well as support for Metal, Apple Accelerate, and Intel MKL for CPU use), you can easily create a REST API OpenAI compatible server or run via Python bindings. Key features include: ✅Prefix caching, continuous batching ✅Flash Attention V2 ✅Device offloading ✅GGUF or Hugging Face models ✅2, 3, 4, 5, 6 and 8 bit quantization ✅X-LoRA MoE non-granular scalings for fast inference ✅Grammar support ✅Continuous batching ✅LoRA support with weight merging ✅LlamaIndex 🦙 integration ...and much more. Incorporation into our GraphReasoning multi-agent modeling framework & LlamaIndex 🦙 allows you to combine in-context learning with adversarial agentic strategies, to dive deep into complex scientific analyses, such as to predict material behaviors, generate hypotheses, analyze papers and data, develop new research concepts, and much more. Check out mistral.​rs: Join our Discord here: Rust Trending Rust Language
0:35

Check out mistral.​rs, our #Rust-based open source inference engine allowing for fast #LLM serving for a variety of architectures including X-LoRA mixture-of-expert (MoE) models, Llama-3, Mistral/Mixtral, Gemma & many others. Built on the Hugging Face #Candle framework for #Rust w/ custom CUDA kernels in the backend (as well as support for Metal, Apple Accelerate, and Intel MKL for CPU use), you can easily create a REST API OpenAI compatible server or run via Python bindings. Key features include: ✅Prefix caching, continuous batching ✅Flash Attention V2 ✅Device offloading ✅GGUF or Hugging Face models ✅2, 3, 4, 5, 6 and 8 bit quantization ✅X-LoRA MoE non-granular scalings for fast inference ✅Grammar support ✅Continuous batching ✅LoRA support with weight merging ✅LlamaIndex 🦙 integration ...and much more. Incorporation into our GraphReasoning multi-agent modeling framework & LlamaIndex 🦙 allows you to combine in-context learning with adversarial agentic strategies, to dive deep into complex scientific analyses, such as to predict material behaviors, generate hypotheses, analyze papers and data, develop new research concepts, and much more. Check out mistral.​rs: Join our Discord here: Rust Trending Rust Language

Markus J. Buehler

73,581 views • 2 years ago

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

103,412 views • 1 month ago

Today, we’re launching Gemma 4, our most intelligent open models to date. Built with the same breakthrough technology as Gemini 3, Gemma 4 brings advanced reasoning to your personal hardware and devices. Here’s what Gemma 4 unlocks for developers: — Intelligence-per-parameter: Our 31B (Dense) and 26B (MoE) models deliver state-of-the-art performance for their size, outcompeting models 20x their size on Arena.ai — Commercial flexibility: Released under a permissive Apache 2.0 license for complete developer flexibility and digital sovereignty — Agentic workflows: Native support for function-calling and structured JSON output allows you to build reliable, autonomous agents — Multimodal edge AI: The E2B and E4B models bring native vision, audio, and low latency to mobile and IoT devices — Long-context reasoning: Up to 256K context windows allow you to process entire repositories or large documents in a single prompt Whether you're building global applications in 140+ languages or local-first AI code assistants, Gemma 4 is built to be your foundation. Explore in Google AI Studio or download the weights on Hugging Face, Kaggle, and ollama.
2:51

Today, we’re launching Gemma 4, our most intelligent open models to date. Built with the same breakthrough technology as Gemini 3, Gemma 4 brings advanced reasoning to your personal hardware and devices. Here’s what Gemma 4 unlocks for developers: — Intelligence-per-parameter: Our 31B (Dense) and 26B (MoE) models deliver state-of-the-art performance for their size, outcompeting models 20x their size on Arena.ai — Commercial flexibility: Released under a permissive Apache 2.0 license for complete developer flexibility and digital sovereignty — Agentic workflows: Native support for function-calling and structured JSON output allows you to build reliable, autonomous agents — Multimodal edge AI: The E2B and E4B models bring native vision, audio, and low latency to mobile and IoT devices — Long-context reasoning: Up to 256K context windows allow you to process entire repositories or large documents in a single prompt Whether you're building global applications in 140+ languages or local-first AI code assistants, Gemma 4 is built to be your foundation. Explore in Google AI Studio or download the weights on Hugging Face, Kaggle, and ollama.

Google AI

1,670,503 views • 5 months ago

Nvidia has just announced Alpamayo 2 Super, an open 34 billion parameter reasoning vision-language-action model designed to accelerate the development of autonomous vehicles. This new model combines the NVIDIA Cosmos 3 Super reasoning model with a 2 billion parameter diffusion-based action expert model, and is post trained with reinforcement learning. The model can return multiple outputs: future trajectory plans, reasoning traces, grounded answers to questions about the scenes, and auto label generation. The model weights are now available for anyone to download on Hugging Face, and the inference code has been posted to GitHub. Distilled models can be deployed commercially without any further permission from Nvidia, and model outputs carry no license conditions. Automakers can distill down a compact version of this model that can run on the Nvidia computer in the car. Major kudos to Nvidia and Jensen Huang for advancing the state of the industry by releasing this as an open model with permissive licensing. Jensen isn't just paying lip service to the idea of open models, Nvidia is actually contributing to the ecosystem — and it's great for their business, because it helps sell more Thor computers that go in the car. Anyone can go download the model and play with it. If you do, let me know what you think. Personally I think it's so cool that we have open weights models that are this advanced, for anyone to download.
2:30

Nvidia has just announced Alpamayo 2 Super, an open 34 billion parameter reasoning vision-language-action model designed to accelerate the development of autonomous vehicles. This new model combines the NVIDIA Cosmos 3 Super reasoning model with a 2 billion parameter diffusion-based action expert model, and is post trained with reinforcement learning. The model can return multiple outputs: future trajectory plans, reasoning traces, grounded answers to questions about the scenes, and auto label generation. The model weights are now available for anyone to download on Hugging Face, and the inference code has been posted to GitHub. Distilled models can be deployed commercially without any further permission from Nvidia, and model outputs carry no license conditions. Automakers can distill down a compact version of this model that can run on the Nvidia computer in the car. Major kudos to Nvidia and Jensen Huang for advancing the state of the industry by releasing this as an open model with permissive licensing. Jensen isn't just paying lip service to the idea of open models, Nvidia is actually contributing to the ecosystem — and it's great for their business, because it helps sell more Thor computers that go in the car. Anyone can go download the model and play with it. If you do, let me know what you think. Personally I think it's so cool that we have open weights models that are this advanced, for anyone to download.

Whole Mars Catalog

45,595 views • 1 month ago

After years of trading other people’s products, I’m finally launching one of my own. Podium is now open to everyone in beta. It’s the first prediction market built natively on Robinhood Chain, and $POD is launching with it. CA: 0x451f509516a6c742bbc7b4869ad8f8de33823929 This one feels different from another trade or another call. I’ve wanted to build something of my own for a long time, and seeing it actually live is pretty surreal. Points are already rolled out too, so you can start using Podium and earning from day one.
0:58

After years of trading other people’s products, I’m finally launching one of my own. Podium is now open to everyone in beta. It’s the first prediction market built natively on Robinhood Chain, and $POD is launching with it. CA: 0x451f509516a6c742bbc7b4869ad8f8de33823929 This one feels different from another trade or another call. I’ve wanted to build something of my own for a long time, and seeing it actually live is pretty surreal. Points are already rolled out too, so you can start using Podium and earning from day one.

qianbaidu.eth 千百度

113,939 views • 28 days ago

Alibaba just dropped Qwen3.5-397B-A17B and there's a lot to unpack. 397B params, 17B active per forward pass. Sparse MoE done right. But the real story isn't the size—it's the architecture choices. The MoE Design Most MoE models feel like bolt-ons. Qwen 3.5's sparse activation is native—only 4.3% of parameters fire per token. That's how you get trillion-parameter-class performance without trillion-parameter inference costs. The 0.8 RMB/million tokens pricing isn't subsidized; it's structurally earned. Native Multimodal, Not Glued-On This is a vision-language model from the ground up. Heterogeneous architecture—separate processing pipelines for text, image, video that fuse early. Not a vision encoder slapped onto an LLM. The result: 90.8 on OmniDocBench, 79.0 on MMMU-Pro. Document understanding and visual reasoning without the usual brittleness. The Context Window Reality Qwen3.5-Plus (the hosted version) ships with 1M tokens by default. That's not a marketing number—they're actually positioning it for long-document workflows. With built-in adaptive tool use, it's clearly aimed at agentic automation, not just chat. What Actually Impressed Me • FP8 native pipeline: ~50% activation memory reduction • Async RL framework for continuous refinement—training and inference workloads separated • 201 languages (up from 119), 250k vocab for better low-resource encoding • Apache 2.0 license. Full weights on HuggingFace and ModelScope. The Benchmark Context 76.4 on SWE-bench Verified puts it in the range where it can handle real debugging workflows. 72.9 on BFCL v4 for agentic tool use. 88.4 on GPQA Diamond. These aren't SOTA in isolation, but the breadth is unusual—strong across reasoning, coding, multimodal, and agentic tasks. The Honest Caveat I haven't stress-tested the 1M context for needle-in-haystack retrieval yet. And "native multimodal" claims need real-world torture testing—PDFs with tables, charts, mixed layouts. Benchmarks are benchmarks. Bottom Line This isn't just another model release. It's a bet on efficient scale: big model capabilities, small active compute, open weights. At 1/18th the cost of Gemini 3 Pro, it's going to force pricing conversations across the board.
1:03

Alibaba just dropped Qwen3.5-397B-A17B and there's a lot to unpack. 397B params, 17B active per forward pass. Sparse MoE done right. But the real story isn't the size—it's the architecture choices. The MoE Design Most MoE models feel like bolt-ons. Qwen 3.5's sparse activation is native—only 4.3% of parameters fire per token. That's how you get trillion-parameter-class performance without trillion-parameter inference costs. The 0.8 RMB/million tokens pricing isn't subsidized; it's structurally earned. Native Multimodal, Not Glued-On This is a vision-language model from the ground up. Heterogeneous architecture—separate processing pipelines for text, image, video that fuse early. Not a vision encoder slapped onto an LLM. The result: 90.8 on OmniDocBench, 79.0 on MMMU-Pro. Document understanding and visual reasoning without the usual brittleness. The Context Window Reality Qwen3.5-Plus (the hosted version) ships with 1M tokens by default. That's not a marketing number—they're actually positioning it for long-document workflows. With built-in adaptive tool use, it's clearly aimed at agentic automation, not just chat. What Actually Impressed Me • FP8 native pipeline: ~50% activation memory reduction • Async RL framework for continuous refinement—training and inference workloads separated • 201 languages (up from 119), 250k vocab for better low-resource encoding • Apache 2.0 license. Full weights on HuggingFace and ModelScope. The Benchmark Context 76.4 on SWE-bench Verified puts it in the range where it can handle real debugging workflows. 72.9 on BFCL v4 for agentic tool use. 88.4 on GPQA Diamond. These aren't SOTA in isolation, but the breadth is unusual—strong across reasoning, coding, multimodal, and agentic tasks. The Honest Caveat I haven't stress-tested the 1M context for needle-in-haystack retrieval yet. And "native multimodal" claims need real-world torture testing—PDFs with tables, charts, mixed layouts. Benchmarks are benchmarks. Bottom Line This isn't just another model release. It's a bet on efficient scale: big model capabilities, small active compute, open weights. At 1/18th the cost of Gemini 3 Pro, it's going to force pricing conversations across the board.

Bo Wang

13,221 views • 6 months ago

QVAC SDK 0.12.0 is now live, bringing longer context, increased memory optimisation, new modalities, and broader ecosystem support directly to your device. Key Features and Updates: - TurboQuant KV-Cache Quantization: Fit much longer context in the same memory. TurboQuant, an algorithm from Google Research, compresses the KV cache by up to 5x, near-lossless. - Text-to-Video: Generate video from a text prompt, fully local, with the new wan2.1 model in the Diffusion addon - Apple Metal Performance for Flux2-klein: Diffusion on Apple Silicon now matches MLX performance, the native benchmark for Apple GPUs - Robot Control (new VLA addon): A GGML-based Vision-Language-Action addon brings fast, efficient robot control to edge devices - Coding Assistant / Harness Support: QVAC now works with OpenCode and OpenClaw as a local provider. A new @qvac/ai-sdk-provider package automates model registry and provider integration - Cross-Platform Voice: Text-to-speech and Parakeet transcription moved from ONNX to the GGML engine for better CPU and GPU support on macOS, iOS, Windows, Linux, and Android. Parakeet also adds long-term streaming diarization (tracking who spoke when on live audio) - Faster Lightweight Visual Classification: A new GGML-based Classification addon delivers millisecond-level classification, useful where a vision-language model (VLM) would be unnecessarily slow - Under the Hood: Fabric synced to llama.cpp v8828 (from v8189), plus GPU acceleration added to image-upscale models for faster results Full release notes:
1:29

QVAC SDK 0.12.0 is now live, bringing longer context, increased memory optimisation, new modalities, and broader ecosystem support directly to your device. Key Features and Updates: - TurboQuant KV-Cache Quantization: Fit much longer context in the same memory. TurboQuant, an algorithm from Google Research, compresses the KV cache by up to 5x, near-lossless. - Text-to-Video: Generate video from a text prompt, fully local, with the new wan2.1 model in the Diffusion addon - Apple Metal Performance for Flux2-klein: Diffusion on Apple Silicon now matches MLX performance, the native benchmark for Apple GPUs - Robot Control (new VLA addon): A GGML-based Vision-Language-Action addon brings fast, efficient robot control to edge devices - Coding Assistant / Harness Support: QVAC now works with OpenCode and OpenClaw as a local provider. A new @qvac/ai-sdk-provider package automates model registry and provider integration - Cross-Platform Voice: Text-to-speech and Parakeet transcription moved from ONNX to the GGML engine for better CPU and GPU support on macOS, iOS, Windows, Linux, and Android. Parakeet also adds long-term streaming diarization (tracking who spoke when on live audio) - Faster Lightweight Visual Classification: A new GGML-based Classification addon delivers millisecond-level classification, useful where a vision-language model (VLM) would be unnecessarily slow - Under the Hood: Fabric synced to llama.cpp v8828 (from v8189), plus GPU acceleration added to image-upscale models for faster results Full release notes:

QVAC

9,932,369 views • 3 months ago

Introducing PhoneLLM, an open model for voice agents. GPT 5.6 Terra performance on typical voice agent tasks at 1/3 the latency and 1/18 the cost. For voice agents, we need models that are both very low latency and very good at tool calling and instruction following. There's a trade-off here, and we often have to compromise on either latency or capability when building voice agents. With PhoneLLM (and the training and data stack that made this model possible) we're fixing this problem. For the last couple of years, most of the effort in frontier model development has gone towards leveraging test-time compute. Which is awesome! Models of all shapes and sizes are available that perform really, really well ... if you have "thinking" turned on for your model. But if you need your agent to respond at voice conversation speed, you can't use thinking models. PhoneLLM is a full-weights fine-tune of NVIDIA Nemotron Nano 30B. We trained on a wide range of real-world telephone and customer support use cases. The training focused on taking the excellent Nano 30B base capabilities and teaching the model to do typical voice agent tasks with thinking disabled. The results are really good: accurate tool calling and concise, on-topic responses in long conversations. And fast: TTFAT measured server-side is <100ms if you run PhoneLLM on a lightly loaded B200. :-) But seriously, when we characterize model latency, we do it with full, end-to-end, batched request simulations using real Pipecat voice agent pipelines. You can serve more than 80 concurrent agents on a single B200 with P95 end-to-end TTFAT <600ms. Including network overhead. That's an LLM cost-per-minute around $0.0025. (1/4 of a cent.) At a latency lower than any third-party API offers today. More details about this model, including weights on Hugging Face, how to spin it up with one click on Modal, and a starter project repo you can clone, are in the thread ...
3:28

Introducing PhoneLLM, an open model for voice agents. GPT 5.6 Terra performance on typical voice agent tasks at 1/3 the latency and 1/18 the cost. For voice agents, we need models that are both very low latency and very good at tool calling and instruction following. There's a trade-off here, and we often have to compromise on either latency or capability when building voice agents. With PhoneLLM (and the training and data stack that made this model possible) we're fixing this problem. For the last couple of years, most of the effort in frontier model development has gone towards leveraging test-time compute. Which is awesome! Models of all shapes and sizes are available that perform really, really well ... if you have "thinking" turned on for your model. But if you need your agent to respond at voice conversation speed, you can't use thinking models. PhoneLLM is a full-weights fine-tune of NVIDIA Nemotron Nano 30B. We trained on a wide range of real-world telephone and customer support use cases. The training focused on taking the excellent Nano 30B base capabilities and teaching the model to do typical voice agent tasks with thinking disabled. The results are really good: accurate tool calling and concise, on-topic responses in long conversations. And fast: TTFAT measured server-side is <100ms if you run PhoneLLM on a lightly loaded B200. :-) But seriously, when we characterize model latency, we do it with full, end-to-end, batched request simulations using real Pipecat voice agent pipelines. You can serve more than 80 concurrent agents on a single B200 with P95 end-to-end TTFAT <600ms. Including network overhead. That's an LLM cost-per-minute around $0.0025. (1/4 of a cent.) At a latency lower than any third-party API offers today. More details about this model, including weights on Hugging Face, how to spin it up with one click on Modal, and a starter project repo you can clone, are in the thread ...

kwindla

324,083 views • 7 days ago

Historically, RL policies for robots have been trained in synthetic, untextured environments, limiting perceptive policies to depth images where the sim-to-real gap is manageable. RGB has always had more potential, but leveraging it to train policies in simulation remained an open problem. Partnering with Niantic Spatial 🌎 and NVIDIA Robotics, we built a pipeline that addresses exactly that. We can now scan a real deployment site with off-the-shelf hardware, reconstruct it into a photorealistic Gaussian splat, and run massively parallel RL training. The policies trained in our Gym environment then transfer zero-shot to the real robot and environments they were trained for. This enables faster deployment of more capable and robust policies for the end user. The new resulting capabilities are a big step towards solving sim-to-real and also apply well beyond navigation. Read the full technical breakdown on our blog; link in the comments. #HumanoidRobots #Flexion #NianticSpatial #NVIDIA
1:21

Historically, RL policies for robots have been trained in synthetic, untextured environments, limiting perceptive policies to depth images where the sim-to-real gap is manageable. RGB has always had more potential, but leveraging it to train policies in simulation remained an open problem. Partnering with Niantic Spatial 🌎 and NVIDIA Robotics, we built a pipeline that addresses exactly that. We can now scan a real deployment site with off-the-shelf hardware, reconstruct it into a photorealistic Gaussian splat, and run massively parallel RL training. The policies trained in our Gym environment then transfer zero-shot to the real robot and environments they were trained for. This enables faster deployment of more capable and robust policies for the end user. The new resulting capabilities are a big step towards solving sim-to-real and also apply well beyond navigation. Read the full technical breakdown on our blog; link in the comments. #HumanoidRobots #Flexion #NianticSpatial #NVIDIA

Flexion

72,096 views • 1 month ago

Anthropic's in trouble, again. The entire Claude experience is now available at 1/6th the price. Kimi now does everything Claude does, powered by K2.6, a 1-trillion-parameter MoE model that activates only 32B parameters per token. It covers all three features Claude has (Chat, Code, and Cowork): 1) Kimi Chat runs in four modes - Instant for fast responses - Thinking for deep reasoning - Agent for multi-step execution - and Agent Swarm for parallel workloads. There's a 262K context window across all of them. 2) Kimi Code is the open-source CLI coding agent with K2.6 as the default backend. K2.6 ranked #1 on OpenRouter's programming leaderboard by weekly usage. 3) Kimi Agent is the Cowork equivalent. It generates: - full websites with database and auth - presentation decks (editable PPTX output) - spreadsheets with formulas and charts - word docs and structured research reports. On top of this, Kimi K2.6 is also trained to decompose tasks into up to 300 parallel sub-agents. This helps it retain coherence even across 4,000+ tool calls in a single run, with sessions sustaining up to 13 hours. On SWE-Bench Pro: - Kimi K2.6 → 58.6 - GPT-5.4 xhigh → 57.7 - Gemini 3.1 Pro → 54.2 - Claude Opus 4.6 → 53.4 Kimi K2.6 model is open weights and self-hostable on 4x H100s in INT4. Find the link to the HuggingFace model page in the replies!
0:41

Anthropic's in trouble, again. The entire Claude experience is now available at 1/6th the price. Kimi now does everything Claude does, powered by K2.6, a 1-trillion-parameter MoE model that activates only 32B parameters per token. It covers all three features Claude has (Chat, Code, and Cowork): 1) Kimi Chat runs in four modes - Instant for fast responses - Thinking for deep reasoning - Agent for multi-step execution - and Agent Swarm for parallel workloads. There's a 262K context window across all of them. 2) Kimi Code is the open-source CLI coding agent with K2.6 as the default backend. K2.6 ranked #1 on OpenRouter's programming leaderboard by weekly usage. 3) Kimi Agent is the Cowork equivalent. It generates: - full websites with database and auth - presentation decks (editable PPTX output) - spreadsheets with formulas and charts - word docs and structured research reports. On top of this, Kimi K2.6 is also trained to decompose tasks into up to 300 parallel sub-agents. This helps it retain coherence even across 4,000+ tool calls in a single run, with sessions sustaining up to 13 hours. On SWE-Bench Pro: - Kimi K2.6 → 58.6 - GPT-5.4 xhigh → 57.7 - Gemini 3.1 Pro → 54.2 - Claude Opus 4.6 → 53.4 Kimi K2.6 model is open weights and self-hostable on 4x H100s in INT4. Find the link to the HuggingFace model page in the replies!

Avi Chawla

110,057 views • 3 months ago

Researchers found a way to make LLMs 8.5x faster! (without compromising accuracy) Speculative decoding is quite an effective way to address the single-token bottleneck in traditional LLM inference. A small "draft" model first generates the next several tokens, then the large model verifies all of them at once in a single forward pass. If a token at any position is wrong, you keep everything before it and restart from there. This never does worse than normal decoding. But current drafters in Speculative decoding still guess one token at a time. That makes the drafting step itself a bottleneck, capping real-world speedups at 2-3x. DFlash is a new technique that swaps the autoregressive drafter with a lightweight block diffusion model that guesses all tokens in one parallel shot. Drafting cost stays flat no matter how many tokens you speculate. On top of that, the drafter is conditioned on hidden features pulled from multiple layers of the target model and injected into every draft layer, so it makes significantly better guesses than a drafter working from scratch. In the side-by-side demo below, vanilla decoding runs at 48.5 tokens/sec. DFlash hits 415 tokens/sec on the same model, with zero quality loss. It's already integrated with vLLM, SGLang, and Transformers, with draft models on HuggingFace for several models like Qwen3, Qwen3.5, Llama 3.1, Kimi-K2.5, gpt-oss, and many more. I have shared the GitHub repo in the replies! KV caching is another must-know technique to boost LLM inference. I recently wrote an article about it. Read it below. 👉 Over to you: What use case are you working on that can benefit from this new technique?
0:26

Researchers found a way to make LLMs 8.5x faster! (without compromising accuracy) Speculative decoding is quite an effective way to address the single-token bottleneck in traditional LLM inference. A small "draft" model first generates the next several tokens, then the large model verifies all of them at once in a single forward pass. If a token at any position is wrong, you keep everything before it and restart from there. This never does worse than normal decoding. But current drafters in Speculative decoding still guess one token at a time. That makes the drafting step itself a bottleneck, capping real-world speedups at 2-3x. DFlash is a new technique that swaps the autoregressive drafter with a lightweight block diffusion model that guesses all tokens in one parallel shot. Drafting cost stays flat no matter how many tokens you speculate. On top of that, the drafter is conditioned on hidden features pulled from multiple layers of the target model and injected into every draft layer, so it makes significantly better guesses than a drafter working from scratch. In the side-by-side demo below, vanilla decoding runs at 48.5 tokens/sec. DFlash hits 415 tokens/sec on the same model, with zero quality loss. It's already integrated with vLLM, SGLang, and Transformers, with draft models on HuggingFace for several models like Qwen3, Qwen3.5, Llama 3.1, Kimi-K2.5, gpt-oss, and many more. I have shared the GitHub repo in the replies! KV caching is another must-know technique to boost LLM inference. I recently wrote an article about it. Read it below. 👉 Over to you: What use case are you working on that can benefit from this new technique?

Avi Chawla

157,390 views • 3 months ago