Anthropic · Model release
Claude Opus 5.5: Anthropic's new default at a lower price than Opus 5
Claude Opus 5.5 replaces Opus 5 as Anthropic's recommended starting point — cheaper per token ($4/$20 vs $5/$25) and at the top of the rebased Artificial Analysis Intelligence Index and CursorBench 4.0.
What changed
- Priced $4/$20 per MTok with cache reads at $0.20/MTok, a 1M-token context and 128K max output (300K on the Batches API behind a beta header); no long-context surcharge.
- Anthropic's model docs now say to start with Claude Opus 5.5 for most workloads, and describe it as close to Fable 5.1 on most work at about 40% lower running cost than Opus 5.
- Leads the Artificial Analysis Intelligence Index v4.3.2 (57.6 at max effort, shown as 58) and CursorBench 4.0 (57.8% at Max, $13.43/task; 56.0% at High for $3.97/task).
Best fit
- Agentic coding
- Knowledge work
- Opus 5 workloads at a lower price
What to be careful about
- Adaptive thinking can no longer be disabled and forced tool use now returns an error — both breaking changes versus Opus 5, so re-test tool-calling code before switching.
- Max effort is verbose: on CursorBench 4.0 the Max row costs over 3x the High row for under two points of extra correctness.
- Sonnet 5.5 and Haiku 5.5 are announced for the coming weeks but not released — do not plan a rollout around them yet.
Dataset snapshot: Claude Opus 5.5 is also represented in the checked comparison dataset. Prices and benchmark figures carry their own field dates.