Qwen on a rampage 🔥 - Qwen3 Coder 480B (35B Active), beats Kimi K2 AND Claude Sonnet 4 - Apache 2.0 licensed! > Upto 1M context > Non-Reasoning > Supports agentic and browser mode So so excited for the smaller models in the series 🤗

Absolutely love playing with this! ❤️

NEW: Onboarding experience on Hugging Face! - Experience the SoTA LLMs and Text to Image models directly 🤩 Really helps contextualise how to use these models and what they are capable of as of today Ofcourse powered by our Inference Providers!

Clarification: This was a cheeky way to say that the latest release from @Alibaba_Qwen sets a new milestone/ hill to climb for rest of the field. We now have an open weights model competing with Opus 4 - the models will only get better and better from here.

This is the dumbest open source models ever will be

Qwen COOKED - beats Kimi K2 and competitive to Claude Opus 4 at 25% total parameters 🤯

missed this, @NVIDIAAIDev silently dropped Open Reasoning Nemotron models (1.5-32B), SoTA on LiveCodeBench, CC-BY 4.0 licensed 🔥 > 32B competing with Qwen3 235B and DeepSeek R1 > Available across 1.5B, 7B, 14B and 32B size > Supports upto 64K output tokens > Utilises GenSelect (combines multiple parallel generations) > Built on top of Qwen 2.5 series > Allows commercial usage Works out of the box in transformers, vllm, mlx, llama.cpp and more!

loosely consider num downloads of openai python package as a leading indicator of AI adoption

It’s the weekend, don’t forget to take it easy, chat!

LLMs but generates only concise and to the point response? Who is building this???