Anthropic — Release Notes

Anthropic’s release notes are per-model and per-API. The behavioral-change disclosures are functionally the known-issues surface: when a model is updated, the notes describe what changes in outputs and what the migration path looks like for teams relying on specific behaviors.

This is the closest the AI tooling space comes to the discipline that PostgreSQL or Python show. Worth tracking if you depend on a specific model’s behavior; “Claude 4.6 → Claude 4.7” can change refusal patterns, output length distributions, and instruction-following in ways that affect production prompts.

The companion to the status page above — the status page is for outages, the release notes are for intentional behavior changes.