Anthropic launched Claude Sonnet 5.5. It scores 56 on the Artificial Analysis Intelligence Index. That is two points behind Opus 5.5. Pricing matches Sonnet 5 at $0.2/$2/$10 per million tokens. Max effort raises its score by 18 points over Sonnet 5. It then uses 193k output tokens per task. Cost per task rises to $7.60 at max effort. It leads on Terminal-Bench 4.0 with 64%. It lags Opus 5.5 on factual knowledge and scientific reasoning. A pre-release bug affected structured outputs. The public release fixes this. Context window stays at 1M tokens. Five effort settings are available.
What changed
Claude Sonnet 5 scored 38 on the Intelligence Index and used fewer output tokens per task.
What it unlocks
Achieving near-parity with Opus 5.5 on agentic and knowledge work benchmarks at max effort.
- 56 Intelligence Index score
- 193k output tokens per task at max effort
- $7.60 cost per task at max effort
- 64% on Terminal-Bench 4.0
What you need to act on it
- max effort setting
- paid API access
Sources