Release 52
Claude Opus 5.5 behind the desk and in Consult
The desk now uses Claude Opus 5.5 wherever a role used Claude Opus 5, except the Quant.
What changed for people using Bellwether
- The Critic reviews on Opus 5.5. Its review after a session and its check before a trade run on Claude Opus 5.5. If that model is unavailable the Critic falls back to Claude Opus 5.
- Opus 5.5 is the fallback. When their own model is unavailable, the Portfolio Manager, the Researcher and Consult now fall back to Claude Opus 5.5 instead of Claude Opus 5.
- Consult's Opus choice is Opus 5.5. The menu under the composer reads Sonnet 5, Opus 5.5, Fable 5.1. A typical Opus 5.5 answer costs about 18 cents.
- The Quant stays on Opus 5. Opus 5.5 passed eight of the Quant's nine test cases on every try and Opus 5 passed nine, so the Quant keeps Claude Opus 5 and Opus 5.5 is not offered for it.
- Earlier answers keep their model. An answer given on Opus 5 still says Opus 5 under it and keeps the price it was charged. The usual cost shown beside Opus 5.5 counts Opus 5.5 answers only.
What you may notice
- On Desk (named "Agent" when this shipped) the Critic's model reads Opus 5.5, and so does the fallback of the Portfolio Manager, the Researcher and Consult. Opus 5.5 joins the models the desk knows: Fable 5.1, Opus 5.5, Opus 5, Sonnet 5.
- Opus 5.5 costs $4 per million tokens in and $20 per million out, a fifth less than Opus 5. Every spend figure uses the price of the model that answered.
- The help pages about the desk name Opus 5.5 where a role moved.
Switched on
- Nothing to switch. The new models apply to every book from this release.
Known gaps and what’s next
- Opus 5.5 stays closed to the Quant until it passes all nine of the Quant's test cases on three tries in a row.
Release 52 e96a96c