Anthropic Releases Claude 3.7 API With Extended Thinking Mode

What Happened

Anthropic has expanded API access to Claude 3.7's extended thinking mode, which was previously available only in limited preview. Extended thinking allows the model to work through problems more slowly and carefully before giving a final answer, similar to how a person might take time to think before responding to a difficult question.

Developers can now control how much thinking time they allocate to the model, balancing speed against accuracy depending on their use case.

Why It Matters

For founders building AI products, this is a practical tool for tasks where accuracy matters more than speed. Think legal document review, financial analysis, complex customer queries, or technical troubleshooting. In these situations, a slightly slower response that is correct is far more valuable than a fast response that might be wrong.

The ability to control thinking time through the API also means you can optimise costs. For simple tasks you can use standard mode. For high-stakes tasks you can dial up the thinking budget and get more reliable results.

For business owners, this matters because it signals that AI tools are becoming more configurable and better suited to professional use cases where errors have real consequences.

What to Do About It

If you are already building with the Claude API, check Anthropic's documentation for how to enable extended thinking in your API calls. It is a relatively straightforward addition to existing integrations.

If you are evaluating AI models for a new project that involves complex reasoning or sensitive decisions, add Claude 3.7 with extended thinking to your testing list. Run it against your specific use case rather than relying solely on benchmark scores, as real-world performance can differ significantly from published numbers.

For non-technical business owners, this feature is most accessible through Claude.ai directly, where you can simply ask for a more careful or thorough response.