Anthropic has released Claude Opus 4.7, a new flagship model showing improved benchmark results for coding and analytical tasks.

Model improvements

Claude Opus 4.7 demonstrates higher benchmark scores and improved performance in code generation, problem analysis, and multi-step reasoning tasks.

The model reportedly sustains longer context handling and applies internal verification steps before producing final outputs to reduce errors.

• Coding: improved generation and debugging capabilities for longer and more complex code fragments.
• Analysis: better performance on tasks requiring multi-step reasoning and structured interpretation of inputs.
• Context: extended context windows allow sustained work on longer documents and codebases.
• Verification: built-in rechecking mechanisms aim to lower the rate of incorrect or inconsistent answers.

API updates and availability

Anthropic also added a new API "thinking" level named xhigh, intended to enable deeper internal reasoning during inference.

This setting is available now in Claude Code, offering developers an option for tasks that benefit from extended deliberation and stability.

Developers and organizations working on longer-form coding tasks and complex analyses may benefit from the model's extended context and verification features.

The updates are delivered through Anthropic's API, allowing integrations into existing developer tooling and automation pipelines.

Anthropic presents Claude Opus 4.7 as a flagship update focused on benchmark gains and developer-facing API controls for coding workflows.