What I learned putting macOS 27's on-device model behind a coding agent
This story is from 2026-10-26. It is preserved in the archive; the latest stories are on the live feed.
In short The context is about 8,000 tokens, shared with the answer, and a log uses more than twice as many tokens per character as prose. Only max_completion_tokens caps the answer, and a cut answer still says "stop" . Asking for a field the text doesn't have can make the answer run away. An anyOf…
Read the full story at DEV Community — AI ↗
Timeline · 1 report
- 2026-10-26 05:28 · DEV Community — AI
What I learned putting macOS 27's on-device model behind a coding agent