Simon Willison ships LLM 0.35 with gpt-6-astra support the same night the model lands
simonw/llm (GitHub)·medium signal
Willison tagged LLM 0.35 at 11:54pm on September 7, adding the `gpt-6-astra` model to the CLI's OpenAI provider. The turnaround matters for anyone scripting evals: `llm -m gpt-6-astra` now works against the same logging, template and fragment machinery as every other model in the tool, so cross-model comparison runs need no new plumbing.