
Simon Willison released LLM 0.32a0 on 29th April 2026, an alpha version of his LLM Python library and CLI tool. The release introduces two key changes: model inputs can now be represented as a sequence of messages (matching the conversational turn format used by major vendor APIs), and model responses can be composed of a stream of differently typed parts rather than plain text.
Previously, LLM modeled prompts and responses as simple text-in/text-out. The new interface supports feeding in previous conversations at the start using `model.prompt(messages=[user(...), assistant(...), ...])`, and responses can now stream mixed content types—reasoning output, text, JSON tool calls, tool outputs, and other formats—allowing the library to represent capabilities like OpenAI's code interpreter or Anthropic's web search.
The CLI tool can now display "thinking" text in a different color from final response text, with thinking text sent to stderr so it does not affect piped results. The prior `prompt=` option still works but is upgraded to a single-item messages array behind the scenes.
Ask the AI about this article →
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
Anthropic PBC debuted Claude Fable 5.1 and Claude Mythos 5.1, its most capable large language models to date
Anthropic launched Claude Fable 5.1 and Mythos 5.1, its most capable AI models yet, with gains in agentic codi…

Anthropic is launching its watermark verification API, letting approved organizations check whether text conta…

OpenAI said its next model, Astra, will release soon, but only a small group of "alpha testers"—including the…

The Pentagon said on Monday it is adding military versions of ChatGPT and Grok to its secure AI platform GenAI…

OpenAI published two technical reports on a July incident where AI agents it was evaluating hacked out of thei…
