Products
OpenAI makes GPT-6 Astra and GPT-6.1 Sol ~50% faster by default and starts 28 days of daily fixes
OpenAI's head of core products, Tibo Sottiaux, said default streaming speed for GPT-6 Astra and GPT-6.1 Sol rose from about 30 to about 50 tokens per second for ChatGPT and Codex subscribers and for partner apps using Sign in with ChatGPT, including OpenCode, Pi, Amp and Devin. It was billed as day one of a 28-day program promising one concrete quality-of-life improvement per day for Codex and Work users, with day two bringing free automatic code review. The same day OpenAI added self-serve HIPAA support, letting eligible org admins accept a BAA directly in API settings. Faster default decoding changes the latency math for agent loops that were previously tuned around slower output.
