Claude Sonnet 5.5
Anthropic premier late-2026 model dominating SWE-bench Verified (98.5%), long-horizon autonomous software engineering, and 92% prompt caching discounts.
| Metric Dimension | Verified Value | Evaluation Benchmark |
|---|---|---|
| Chatbot Arena Elo | 1415 pts | LMSYS Live Battles |
| Coding Index | 98.5 / 100 | SWE-bench Verified |
| Agentic Tool-Use | 98 / 100 | Multi-Turn Tool Calls |
| Input Rate | $2.00 / 1M | Commercial Token API |
| Output Rate | $10.00 / 1M | Generation Billing |
| Context Horizon | 500k | Maximum Token Window |