Codex 28-day program, Day 1: GPT-6 Astra and GPT-6.1 Sol run about 50% faster on subscriptions

#What changed

On Oct 5, Tibo (@thsottiaux) posted Day 1 of the Codex 28-day ship-or-reset program: OpenAI raised the default speed of GPT-6 Astra and GPT-6.1 Sol by about 50% for subscription usage. In a follow-up reply he put it at roughly 50 tokens per second, up from about 30, and added that the models' tokenizer and token efficiency mean tasks need fewer tokens overall. This is the program's first daily ship, not a usage reset.

#Who is affected

People using GPT-6 Astra or GPT-6.1 Sol through a ChatGPT subscription, across OpenAI's own products and partner tools that use Sign in with ChatGPT. He named OpenCode, Pi, Amp, and Devin. He did not mention API usage billed per token.

#What to do now

Nothing to change. He said no settings or client updates are needed and the speedup should be felt within about two hours of his 10:20 AM PT post. If you had been moving long jobs to another model only because Sol or Astra felt slow, time a typical run again this afternoon to see whether that still makes sense.

#What is not confirmed

OpenAI has not published benchmarks or a status note, and the 30 and 50 tokens-per-second figures are Tibo's. Dotsfeed has not measured the change. It is not stated whether it applies equally to every plan, region, or reasoning level, or whether the API is affected.