Claude Sonnet 5.5 Rivals Opus With 30% Speed Gain

Anthropic picked a rough day for OpenAI's biggest showcase of the year. Hours before the company's flagship event, Anthropic pushed out Claude Sonnet 5.5, a mid-tier model that runs 30% faster than its predecessor and, on several benchmarks, closes in on Anthropic's own top-shelf Opus. The timing wasn't subtle.

Last week Anthropic released Opus 5.5. This week it's Sonnet's turn, and the gap between the two tiers has gotten noticeably smaller.

That's the real story here: Anthropic isn't just chasing GPT and other frontier models anymore, it's competing with itself. Sonnet 5.5 keeps the same pricing as Sonnet 5, but Anthropic says the new version handles jobs for up to 30% less and matches its predecessor's best output on low or medium effort settings for roughly a tenth of the cost. On Artificial Analysis's Intelligence Index, Sonnet 5.5 scored a 56, trailing only Opus 5.5 among Anthropic's own lineup while beating out competing models from other labs.

Coding results tell a similar story, with Sonnet pulling ahead of Opus on some tests. What Anthropic is claiming about where that improvement is coming from is worth a closer look.