Evidence record / claude-code
A dated record of one change, kept so the writing that cites it can be checked. Compiled from the sources listed below by the research run, not written for reading. The judgment lives in the issues that cite it, below.
2026-10-13-claude-code-safety-related-model-switches-can-override-a-pinned-opus-model
Safety-related model switches can override a pinned Opus model. A pinned Opus configuration no longer chooses the replacement model after a safety flag on the Anthropic API; the API selects one for the type of flag. Effort survives the switch from 2.1.287. Pinning protects an evaluation from alias drift, but does not guarantee every request uses the same model.
Channel: tagged-release (on stable via 2.1.285 and 2.1.287). Half: both. Date: 2026-09-28 (2.1.284), 2026-10-01 (2.1.287).
Operator consequence: Keep full IDs for alias stability and record the model actually returned for evaluated turns. Separate flagged switches from ordinary samples before attributing changed results or cost to a pinned model. This is a server-directed exception to model choice, not a general removal of pins.
Receipt
Finding metadata
Run: 2026-10-13-profile-refresh-2026-09-21_2026-10-13-frontier-v0
Finding ID: 2026-10-13-claude-code-safety-related-model-switches-can-override-a-pinned-opus-model
Profile citations
- Claude Code / claim / safety-switch-overrides-opus-pin
- Claude Code / posture / governance
Source links
Primary links, including exact changelog lines when available.
Versioned source: run artifact