Red Hat AI gives teams a place to experiment with newly released models, including Mistral 3, on the day they arrive. Our Day Zero guide shows how to run Mistral 3 today using the Red Hat AI Inference Server and Red Hat OpenShift AI: https://t.co/Sr79WfbrPw Happy experimenting!

Many more details are in our quick-start blog:

The workflow is simple and fast. Load your model, apply the AutoRound modifier, compress the model, and serve it in @vllm_project. Support for Llama, Qwen, and other open-weight LLMs means you can start experimenting right away.

I can personally vouch for a dozen of these - from the genius silverware storage, to the perfect leggings with 70+ reviews, to the hands-down best pillows on earth.

AutoRound learns how each tensor should round and clip for the best quality possible. This delivers standout low bit performance in formats like W4A16.

Red Hat AI and @intel have teamed up to bring a major upgrade to low bit LLMs. AutoRound is now integrated directly into LLM Compressor, giving developers a powerful way to shrink models, boost speed, and keep accuracy. And it runs smoothly with @vllm_project. A quick 🧵:

Low-bit LLM quantization doesn’t have to mean painful accuracy trade-offs or massive tuning runs. Intel's AutoRound PTQ algorithm is now integrated into LLM Compressor, producing W4A16 compressed-tensor checkpoints you can serve directly with vLLM across Intel Xeon, Gaudi, Arc

Thanks to the great collaboration with @vllm_project , LLM Compressor team, and @RedHat_AI team for making this happen! If you want a model with smaller-size, high accuracy, and deployed w/ vLLM, AutoRound is your best choice. 🌟 and give a try:

If you built agents with Llama Stack's original Agent APIs, you've probably seen that they are being deprecated in favor of the OpenAI compatible Responses API. Migrating does not require starting over. There are two practical paths you can take. Approach 1 is a

What if your money could do more than grow? Robin John’s new book shows you how your investments can honor God and serve your neighbor. Join the Good Investor Movement and start investing with purpose!

🎉 𝗠𝗶𝗹𝗲𝘀𝘁𝗼𝗻𝗲 𝘂𝗻𝗹𝗼𝗰𝗸𝗲𝗱! The 𝘐𝘯𝘧𝘦𝘳𝘦𝘯𝘤𝘦𝘖𝘱𝘴: 𝘚𝘵𝘢𝘵𝘦 𝘰𝘧 𝘵𝘩𝘦 𝘔𝘰𝘥𝘦𝘭 𝘚𝘦𝘳𝘷𝘪𝘯𝘨 𝘊𝘰𝘮𝘮𝘶𝘯𝘪𝘵𝘪𝘦𝘴 newsletter from @RedHat_AI just reached 𝟭,𝟬𝟬𝟬 𝘀𝘂𝗯𝘀𝗰𝗿𝗶𝗯𝗲𝗿𝘀 in only 5 months! 🚀 A huge thank-you to everyone who’s

Are you running vLLM on Kubernetes and tired of guessing concurrency thresholds? This new Red Hat article walks through how to autoscale vLLM on OpenShift AI using real service metrics instead of generic request counts. KServe and KEDA work together to scale GPU model servers

Our colleague @dougbtv built a very cool integration that shows what is possible with vLLM-Omni, a new multimodal framework introduced last week. He connected vLLM-Omni to @ComfyUI, a community favorite for designing advanced diffusion workflows. The project includes a new

📢vLLM v0.12.0 is now available. For inference teams running vLLM at the center of their stack, this release refreshes the engine, extends long-context and speculative decoding capabilities, and moves us to a PyTorch 2.9.0 / CUDA 12.9 baseline for future work.

Salt? Slush? Holiday road trips? 🎄 Slick Mist® makes cleanup fast and satisfying. Perfect for anyone who hates winter grime.

It may be my first time at #AWSreInvent but it sure isn’t for @RedHat! We’re out here in Vegas running live demos all week, including my favorite Blackjack + AI game powered by @vllm_project for model inference ⚡️ and #ModelContextProtocol Agents 🤖

It’s @vllm_project party time at NeurIPS!

Red Hat's expanded collaboration with @awscloud is empowering IT decision-makers to run high-performance, efficient #AI inference at scale with @RedHat_AI. Check it out. #AWSreInvent

Our latest PyTorch Foundation Spotlight features @RedHat's Joseph Groenenboom and Stephen Watt on the importance of optionality, open collaboration, and strong governance in building healthy and scalable AI ecosystems. In this Spotlight filmed during PyTorch Conference 2025,

Christianity and classical liberalism—can they coexist? Carl Trueman and Vincent Phillip Muñoz debate the biggest questions facing American conservatism. Christian nationalism. Post-liberalism. Government's role in virtue. Where nationalism becomes dangerous. FULL

🎉 Congratulations to the Mistral team on launching the Mistral 3 family! We’re proud to share that @MistralAI, @NVIDIAAIDev, @RedHat_AI, and vLLM worked closely together to deliver full Day-0 support for the entire Mistral 3 lineup. This collaboration enabled: • NVFP4

Congrats to @MistralAI on launching the Mistral 3 family under the Apache 2.0 license. We worked together to enable upstream @vllm_project support and collaborated on creating the FP8 and NVFP4 Mistral Large 3 checkpoints through llm-compressor for efficient deployment. 🚀

We @RedHat_AI have partnered with Mistral to make Mistral Large 3 more accessible to the open-source community. High-quality FP8 and NVFP4 mdls, built with our llm-compressor! Expect models that are 2–3.5x smaller with competitive accuracy across a wide range of evals.

Explore some of OpenShift #AI's capabilities for scaling LLM model servers with #KServe and #vLLM. While autoscaling has its limitations, it can be a valuable tool for an IT team trying to optimize the costs of the models they are serving. https://t.co/k57HhY2Iyp