Anthropic has released Claude Sonnet 5.5, the second entry in its 5.5 model generation, arriving six days after Claude Opus 5.5. Where Opus 5.5 is pitched as the flagship, Sonnet 5.5 is built for the bulk of day-to-day work: bug fixes, routine coding, and document or spreadsheet tasks that need speed as much as reasoning depth.
The headline number is a 70.6% score on Terminal-Bench 4.0, a test of agentic command-line coding skill. That is a sharp jump over the previous Sonnet generation on the same benchmark and edges past Opus 5.5's reported 66.4%, though Anthropic measured the two models under different reasoning-effort settings, so the comparison isn't exact.
What stands out is the pricing. Sonnet 5.5 keeps the same $2 per million input tokens and $10 per million output tokens that Sonnet 5 charged, rather than rising with the new generation, as some rivals have done on their latest models. Cached input tokens cost a tenth of that rate, at $0.20 per million, which matters for workloads that repeat large prompt prefixes.
Sonnet 5.5 carries a one-million-token context window and can return up to 128,000 tokens in a single response. Thinking is adaptive and switched on by default, and the model keeps vision, tool calling and structured JSON output from earlier releases. Anthropic has said Claude Haiku 5.5 is due in the coming weeks, which would complete the 5.5 line-up across all three sizes.
Source: MarkTechPost
