Claude Sonnet 4.6: the 'middle' model that beats Opus 4.5 in coding
In one sentence Anthropic releases Sonnet 4.6: 79.6% on SWE-bench Verified, 72.5% on OSWorld-Verified (on par with Opus 4.6), better prompt-injection resistance. Pricing unchanged at $3/$15.
Sonnet is the middle model in the Claude family: cheaper than Opus, more capable than Haiku. Version 4.6 lands on February 17, 2026 and does something unusual: in coding, testers prefer it to Opus 4.5 (the previous "premium" model) 59% of the time. For price-to-performance, it's the default pick.
Concrete numbers: 79.6% on SWE-bench Verified (the most-used benchmark measuring how well an AI fixes real GitHub bugs), and 72.5% on OSWorld-Verified (computer use, i.e. how well it clicks, scrolls, navigates a real app) — basically on par with the flagship Opus 4.6.
What it means in practice: pricing unchanged at $3 input / $15 output per million tokens, 1M context in beta, and noticeably better prompt-injection resistance (important for agents handling potentially hostile email or documents). It's immediately the default on Claude.ai for Free and Pro plans.
Companies
Anthropic
Tools
Claude Sonnet 4.6
Tags
Sources