Anthropic launches Claude Sonnet 4.6 with 1M-token context and broader coding gains

Anthropic announced Claude Sonnet 4.6 on February 17, 2026 as its newest Sonnet-tier model. The company positions it as a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design, while keeping API pricing at $3/$15 per million tokens. Anthropic also made Sonnet 4.6 the default model for Free and Pro users in Claude.ai and Claude Cowork.

The headline capability change is a 1M token context window in beta. Anthropic says that is enough for entire codebases, long contracts, or dozens of research papers in one request, and it frames the release as a step toward longer-horizon agent work rather than a narrow benchmark bump. The company also highlighted stronger prompt injection resistance for computer use than Sonnet 4.5.

• 1M token context window in beta
• Same API pricing as Sonnet 4.5 at $3/$15 per million tokens
• Immediate rollout as the default model for Free and Pro users in Claude.ai and Claude Cowork
• Broader platform updates around adaptive thinking, extended thinking, context compaction beta, and general availability for code execution, memory, and programmatic tool calling

Anthropic's own early testing says Claude Code users preferred Sonnet 4.6 to Sonnet 4.5 roughly 70% of the time and even preferred it to Opus 4.5 59% of the time. Those figures are company-reported rather than independently verified, but they matter because they suggest Anthropic is pushing the cheaper Sonnet tier into workloads that previously required an Opus-class model.

For the market, the release matters because Sonnet is Anthropic's volume model tier. If a lower-cost model can handle bigger codebases and more reliable computer-use flows, teams building agents may be able to shift more production traffic away from higher-cost frontier models. The next question is whether customer evals and external benchmarks validate Anthropic's claim that Sonnet 4.6 meaningfully narrows the gap with Opus while preserving strong safety behavior.