Willison's LLM tool adds GPT-6 Astra support

Simon Willison tagged version 0.35 of his LLM command-line tool at 11:54pm on September 8, the same night OpenAI's Astra model went live. The release adds gpt-6-astra to the tool's OpenAI provider. Any script that already calls LLM against other configured models can call Astra the same way, with no new integration code.

Because Astra runs on OpenAI's existing API shape, Willison only needed to add the model ID to the existing provider. No new plugin, no separate wiring. gpt-6-astra now runs through the same logging, prompt templates, and fragment system every other model in the tool uses.

For anyone who scripts model comparisons against LLM, typing llm -m gpt-6-astra now works out of the box. Astra's output logs next to every other model already configured in the tool.

The release notes don't say whether gpt-6-astra needs different handling than other OpenAI models in the tool. They say only that the model ID now works.