Anthropic launched Claude 3.7 Sonnet with a toggle for near-instant responses or extended step-by-step thinking.

The model provides a single endpoint for both standard inference and deep reasoning tasks.

This hybrid approach allows developers to manage latency and compute costs within a single model architecture.

• Jun 06, 2026Anthropic updated prompt caching to automatically read from the longest previously cached prefix when breakpoints are set.
• Jun 06, 2026Anthropic optimized prompt caching support in the Message Batches API to improve cache hit rates.
• Jun 06, 2026Anthropic added a delete endpoint to the Message Batches API for managing batch processing tasks.

This is the world we build in. It moves this fast. We move with it.