Anthropic Releases Claude Sonnet 4.6 as New Default Model With 1M Token Context

A Major Upgrade Across the Board

Anthropic launched Claude Sonnet 4.6 on February 17, 2026, delivering comprehensive improvements across coding, computer use, long-context reasoning, agent planning, and design. The new model immediately became the default for Free and Pro plan users on claude.ai and Claude Cowork.

Performance Highlights

• Coding: Developers preferred Sonnet 4.6 over Sonnet 4.5 approximately 70% of the time in Claude Code testing; it even beat Opus 4.5 in 59% of comparisons
• Computer Use: Achieved human-level performance on complex OSWorld tasks including multi-step spreadsheet navigation and form completion
• Instruction Following: Improved consistency, reduced hallucinations, and better long-horizon planning
• Prompt Injection Resistance: Enhanced defenses compared to prior versions

1M Token Context Window in Beta

Sonnet 4.6 ships with a 1 million token context window in beta. Context Compaction—automatic summarization to extend effective context length—is also available in beta. The model also supports Adaptive Thinking and Extended Thinking.

Pricing and Availability

Pricing remains unchanged from Sonnet 4.5 at $3/$15 per million tokens (input/output). Available via the Claude API, claude.ai, Claude Cowork, Claude Code, and Excel integration as claude-sonnet-4-6.

Safety

Anthropic's extensive safety evaluations found Sonnet 4.6 to be at least as safe as prior models, with strong safety behaviors and no signs of high-stakes misalignment concerns.

Source: Anthropic announcement