Anthropic · Claude 5
Claude Sonnet 5.5
Anthropic's current Sonnet (Sep 2026), scoring within a few points of Opus 5.5 on agentic coding and knowledge work at half the token price.
- AA Intelligence
- 56#2 of 143
- Input / output per 1M
- $2.00/ $10.00
- Output speed
- —
- Context window
- 1M tokens
Composite indexes
- 5647
Leader: Claude Opus 5.5 at 58
Coding
- SWE-bench Pro#2 of 3281.3%
Leader: Claude Opus 5.5 at 89.9%
- Terminal-Bench 4.0#1 of 8970.6%
Leader: Claude Sonnet 5.5 at 70.6%
Agents and tool use
- GDPval-AA#2 of 1041844
Leader: Claude Opus 5.5 at 1846
- OSWorld-Verified#7 of 1780.1%
Leader: Kimi K3 at 84.8%
Reasoning and knowledge
- Humanity’s Last Exam#1 of 14864.5%56.9%
Leader: Claude Sonnet 5.5 at 64.5%
Other published results
| Result | Score | Reported by | Source |
|---|---|---|---|
| SWE-bench Multilingual · max effort, avg of 5 trials | 90.3% | vendor | Claude Sonnet 5.5 System Card ↗ |
| SWE-bench Multimodal · max effort, avg of 5 trials | 54.3% | vendor | Claude Sonnet 5.5 System Card ↗ |
| DeepSWE v1.1 · max effort, avg of 5 trials | 71.0% | vendor | Claude Sonnet 5.5 System Card ↗ |
| FrontierCode v1.1 (Main) · xhigh effort, run by Cognition (46.2% at max) | 52.1% | vendor | Claude Sonnet 5.5 System Card ↗ |
| Terminal-Bench-Science 0.1 · max effort | 59.9% | vendor | Claude Sonnet 5.5 System Card ↗ |
| CursorBench 4.0 · max effort, run by Cursor | 55.5% | vendor | Anthropic: Introducing Claude Sonnet 5.5 ↗ |
| AA-Briefcase v1.1 · max effort; run by Artificial Analysis on a pre-release deployment | 1811 Elo | independent | Claude Sonnet 5.5 System Card, section 8.14.4 ↗ |