Core capabilities
- 1M context and 128K maximum output
- Adaptive thinking with controllable effort
- Strong coding, visual understanding and tool use
- Managed availability across clouds with lower cache-read cost
← Model indexMODEL PROFILE / claude-sonnet-5-5
Anthropic / current
Claude Sonnet 5.5
The faster, lower-cost Claude 5.5 tier for coding and enterprise agents.
Context1M
ModalityText · Image
AccessAPI · Cloud
ReasoningAdaptive
Where it fits.
Complements Opus 5.5 with materially higher speed and lower cost, making it the practical primary model for most agent work. Released September 28, 2026, Sonnet 5.5 runs over 30% faster and costs up to 30% less than Sonnet 5, with cache reads later halved to $0.10 per million.
Best-fit workflows
- Primary coding agent
- Enterprise assistants
- Data and document workflows
04 / OPERATIONSFrom evaluation
From evaluation
to production.
Deployment notes
Start new work here and escalate to Opus 5.5 or Fable 5.1 only where evals show a real gain. Keep tool schemas narrow and return thinking blocks unchanged.
Watch before choosing
Adaptive thinking is on by default; think-block handling differs from Haiku's manual extended thinking. Model switches inside an active conversation need careful migration of prior thinking blocks.
05
Family & versions
- Claude Opus 5.5Flagship
- Claude Fable 5.1Coding flagship
- Claude Sonnet 5.5Balanced
- Claude Haiku 5.5Fast tier
Specifications and positioning checked against provider documentation.
Open official source↗