Model Capability Discovery

Anthropic has enhanced the Models API with new metadata fields that enable runtime discovery of model capabilities. The GET /v1/models and GET /v1/models/{model_id} endpoints now return:

• max_input_tokens: maximum input context window
• max_tokens: maximum output generation tokens
• capabilities: object describing supported features

This allows developers to programmatically query available model capabilities rather than hardcoding assumptions, improving compatibility with future model updates.

Extended Thinking Display Control

The extended thinking feature now supports a new display field that gives developers control over thinking content visibility. By setting thinking.display: "omitted", developers can:

• Receive thinking blocks with empty thinking fields
• Preserve the signature for multi-turn conversation continuity
• Enable faster streaming without thinking content overhead
• Maintain unchanged billing for thinking token usage

This is particularly useful for applications that need extended thinking's reasoning benefits without exposing the internal thought process to end users.

Context

These updates represent incremental improvements to Claude's platform developer experience, reducing friction when building multi-model applications and optimizing response delivery for consumer-facing interfaces.