Anthropic Presents Claude Sonnet 4.6 with Context of Up to 1 Million Tokens and Improved Logic

|
Anthropic Presents Claude Sonnet 4.6 with Context of Up to 1 Million Tokens and Improved Logic

Anthropic has introduced an updated version of its AI model, Claude Sonnet 4.6, which serves as the foundation for the Claude and Claude Cowork services. The new version features significant enhancements in programming, data analysis, task planning, and software interaction. With an expanded context window of up to 1 million tokens (currently in beta), the model can process large data sets, entire codebases, lengthy contracts, and extensive research materials.

This is reported by Finway

Key Advantages of Claude Sonnet 4.6

  • The model demonstrates improved accuracy in following instructions and a reduced number of “hallucinations.”
  • Stability in executing multi-step tasks has been enhanced, which is crucial for implementing complex business processes.
  • Progress has been noted in “computer usage” skills: the AI can interact with software interfaces through virtual clicks and text input, increasing automation capabilities.

The cost of using Claude Sonnet 4.6 remains at the same level as the previous version, ranging from $3 to $15 per million tokens. The model has approached Opus-class quality while providing a more accessible price, expanding its applicability for corporate tasks and developers.

“Sonnet 4.6 has a context window of up to 1 million tokens (in beta). According to the developers, this allows for processing large codebases, lengthy contracts, and sets of research materials in a single request.”

Testing, Security, and Implementation

In tests and initial implementations, Claude Sonnet 4.6 showed increased efficiency in document analysis, interface development, and financial calculations. The model also demonstrated significant advantages in executing agent tasks. Developers with early access more frequently chose Sonnet 4.6 over the previous version, and in some cases, even preferred it to Opus 4.5.

Anthropic conducted extensive security checks on the new model. According to the test results, Claude Sonnet 4.6 demonstrates a reliability level that matches previous versions, and in some aspects, even exceeds them.

Comparison of Claude Sonnet 4.6 with leading AI models based on key benchmarks. Data: Anthropic.