Claude Managed Agents dynamic workflows are now available in public beta. It's a new type of multiagent orchestration, built for your most ambitious workloads. A lead agent writes a plan that runs across many agents in phases, combining the results at the end.

vLLM now supports NVIDIA Vera Rubin. The early results show more than 7.8x the throughput of GB200 on MiniMax M3 on AgentX. @inferact, @NVIDIAAI, @RedHat_AI, and the vLLM community have been bringing vLLM up on Rubin since it was announced. Here is where things stand. 🧵 1/5

Your next meeting comes with a note-taker who can help with the follow-up. The Meetings plugin takes notes for you and saves a personalized summary and next steps in ChatGPT Space, based on what ChatGPT knows about you and the work you’ve done together. Keep notes private or share them with your team, then ask ChatGPT to update a project plan or draft a follow-up. Available in beta for Pro and Business users in the ChatGPT desktop app on macOS. Enterprise is coming soon. Download the desktop app, then search for “Meetings” in the plugin directory.

We shipped four things that were deemed good to great and some math proofs, but the vote is clear and the community demands a reset. I did calibrate it and it *seems* that the game is rigged in reset's favor, but such are the rules at the moment. Therefore ... the reset has been processed. Enjoy!

Day 1/ We have optimized the default speed to be ~50% faster across GPT-6 Astra and GPT-6.1 Sol through the subscription across all our products and partners using Sign in With ChatGPT (including OpenCode, Pi, Amp, Devin, ...). No changes needed on your end and this should be felt within the next two hours.

OpenAI is going to win. Sol 6.1 is frustratingly slow. Astra 6 is one of the most usage hungry models. And yet, subscriber growth in Codex continues to go up according to the OpenAI team. If users are still buying the product even when there are much better alternatives OpenAI clearly has the best brand. I have no loyalty and will always jump to the best products but the normal user apparently wants ChatGPT regardless.

Scheduled Tasks on @Muse - An update on reliability We deployed a fix to Muse to improve scheduled task and cron recovery and reliability especially, and have hardened our monitoring and alerting of the cron systems in the VMs. Over the last few days the Muse team has been deploying a few changes to: Improve the KV cache hit rate after increasing the context window size of the underlying model Improve our VM scheduling algorithm to reduce downtime during deployments Deploy a new authentication token format across the fleet for more fine-grained access. During these changes we had a few SEVs (outages) where the inference tier was unresponsive, and the VM scheduling resulted in VMs not being available for longer periods of time during deployment and being unresponsive to clients. We observed after we recovered from these outages that several crons and scheduled tasks failed to resume / heal properly when the outage was done, resulting in several users receiving reports from their Muse that the crons have been failing for a while. We apologise for this incident, and know how much this disrupts your flow especially since many users rely on muse to reliably deliver outcomes for them. We take these outages seriously, and are running a full audit of our recovery systems and tools, and are adding more dashboards and alerts so you can continue to receive the high quality service we want all users to have.