Anthropic Unveils Claude Sonnet 4.6 with Enhanced Coding Capabilities and 1M Token Context Window

17.02.2026

Anthropic has officially launched Claude Sonnet 4.6, the latest iteration of its mid-tier language model, maintaining the company's consistent four-month release cadence. The update brings significant enhancements across multiple domains, particularly in coding performance, instruction adherence, and computer interaction capabilities.

The new model will serve as the default option for users on Free and Pro subscription tiers. A key highlight of this release is the beta implementation of a 1 million token context window, representing a twofold increase compared to the previous maximum capacity available for Sonnet models. According to Anthropic, this expanded context window provides sufficient capacity to process entire codebases, extensive legal contracts, or multiple research papers within a single request.

This deployment follows closely on the heels of the Opus 4.6 release from two weeks prior, with the updated Haiku model anticipated to arrive in the near future, completing the model family refresh cycle.

Benchmark Performance:

Sonnet 4.6 has achieved notable results across industry-standard evaluation metrics:

• OS World benchmark for computer use capabilities
• SWE-Bench for software engineering tasks
• 60.4% score on ARC-AGI-2, which evaluates cognitive abilities typically associated with human intelligence

While the ARC-AGI-2 performance positions Sonnet 4.6 above most comparable models in its class, it remains behind frontier models including Opus 4.6, Gemini 3 Deep Think, and certain optimized variants of GPT 5.2.

Source:
Anthropic Official Announcement

Tags: AI models Claude Anthropic LLM benchmarks

Share: VK Telegram Twitter