dcdpr / jp / 24529090111

67%

push

github

feat(llm, anthropic): Add Claude Opus 4.7 and `xhigh` effort support (#569)

Add support for the new `claude-opus-4-7` model and introduce `xhigh`
(extra-high) effort level for adaptive thinking models that support it.

`ReasoningDetails::Adaptive` gains a new `xhigh: bool` field to track
per-model support for the `xhigh` effort level. The `adaptive()`
constructor is updated accordingly. Opus 4.7 supports both `xhigh` and
`max`, while Opus 4.6 and Sonnet 4.6 support only `max`.

When a model supports `xhigh`, `ReasoningEffort::XHigh` (and
`ReasoningEffort::Max` as a fallback) now maps to `Effort::XHigh`
instead of falling through to `Effort::High`.

The `context-1m-2025-08-07` beta header gating is also removed: affected
models now unconditionally advertise a 1M context window, removing the
dependency on `BetaFeatures` from `map_model`.

Signed-off-by: Jean Mertz <git@jeanmertz.com>

13 of 26 new or added lines in 2 files covered. (50.0%)

22693 of 35360 relevant lines covered (64.18%)

224.09 hits per line