Anthropic Unveils Claude Sonnet 4.6 With 1 Mn-Token Context and Stronger Coding, Agent Skills
Sonnet 4.6 is its most capable Sonnet model to date and is now the default model across Claude’s free and paid tiers.
Anthropic has introduced Claude Sonnet 4.6, the latest version of its mid-tier AI model, touting major gains in coding, long-context reasoning, computer use and agentic planning as competition among frontier AI systems intensifies.
The company said Sonnet 4.6 is its most capable Sonnet model to date and is now the default model across Claude’s free and paid tiers, with pricing unchanged from Sonnet 4.5.
The model includes a beta 1-million-token context window and improved instruction-following and consistency, allowing performance on some complex tasks that previously required higher-end models.
Anthropic highlighted stronger computer-use abilities, enabling the model to interact with software interfaces in ways similar to a human user, such as navigating spreadsheets or completing multi-step web tasks.
The company said the model also shows improved resistance to prompt-injection attacks and has undergone extensive safety testing, with researchers finding it as safe as or safer than recent Claude releases.
Early enterprise testers reported notable gains in coding, document analysis and multi-step workflows. Some said Sonnet 4.6 approaches higher-tier “Opus” model performance at a lower cost, particularly for tasks such as bug detection, financial analysis and complex agent orchestration.
Anthropic said the model supports features including extended reasoning, tool use, code execution and automated context compaction. Sonnet 4.6 is available through Claude’s web interface, API and major cloud platforms, as the company continues to position its models for enterprise productivity and agent-based automation use cases.
