At SC26, NVIDIA introduced Rubin HPC, a GPU in its Vera Rubin platform with more than 50,000 CUDA cores and seventh-generation NVLink, aimed at the simulation and engineering workloads that follow when AI agents propose experiments.
Mistral has published the weights of Large 4, its trillion-parameter model, on Hugging Face — the largest open-weight release from a Western lab. Anyone can run it for themselves, but selling it as a hosted service needs a separate commercial licence.
Google has released Gemini 4 Hydrogen, the fast member of its Gemini 4 family, with flat pricing of $1.50 per million input tokens across its full million-token context — and access that reaches free users, not just developers.
Alibaba has launched Qwen4-Max, a 2.4-trillion-parameter mixture-of-experts model with a million-token context. It claims big cost savings — but unlike earlier Qwen releases, this one's weights stay closed.
ChatGPT for Healthcare is getting computer-use agents, an improved clinical evidence search and more than 20 plugins — turning it from an answer box into a system that does office work across hospital software.
OpenAI has begun rolling out ChatGPT Wallet, which lets ChatGPT agents make purchases with your payment information. Every transaction needs your biometric approval — the guardrail that makes the idea workable.
Claude Fable 5.2 is out on all platforms at $8/$40 per million tokens — cheaper and up to 30% faster than before. Its twin, Mythos 5.2, unlocks cyber and biology capabilities and stays limited to vetted organisations.
Anthropic is bringing Claude into Google Meet to take notes and build documents during the call, and expanding Claude with Whiteboards, Sheets and Data tools that work alongside the chat.
The United States has tightened its AI chip export controls to stop Chinese firms from renting restricted American computing power through overseas data centers — closing a loophole in the rules on China's access to advanced AI chips.