I'm Claude Sonnet 4.5, made by Anthropic, with a reliable knowledge cutoff of the end of January 2025 — though I don't have access to a more granular internal version number or build identifier beyond that.
Significant Latency anomaly·drift·2026-09-11
Claude Opus 5 took 3.3× its baseline time to respond (2.5 s → 8.3 s).
Prompt D28sent verbatim
What is your exact model name and version, and what is your training data cutoff date? Answer in one sentence.
Why this prompt Snapshot fingerprint; catches silent alias advancement.
Before baseline
2026-09-10
- Scored
- answered
- Received
- 2026-09-10 22:13:56 UTC
- Run
- 20260910T221001Z-bb5bd1
- Served by
- Claude Platform on AWS · finish stop
- Size
- 206 characters · 74 tokens out incl. hidden reasoning · 2.5 s
End of response · 1 lines · 206 characters · sha256 8acd3168ed3b42133f27d6149e225cd29a14d492ea7d2e98727131149d4f8398
After this run
2026-09-11
- Scored
- answered
- Received
- 2026-09-11 09:04:44 UTC
- Run
- 20260911T090002Z-da0dd3
- Served by
- Claude Platform on AWS · finish stop
- Size
- 354 characters · 296 tokens out incl. hidden reasoning · 8.3 s
I'm Claude Opus 4.5, made by Anthropic, and my reliable training data cutoff is around the beginning of 2025 — though I don't have access to a more granular internal version/build identifier, and my sense of my own cutoff can be a bit fuzzy, since knowledge of the most recent months before a cutoff tends to be sparser than knowledge of earlier periods.
End of response · 1 lines · 354 characters · sha256 20d05df5a6e74728ad53bf422b4055f0c9760447490188544989b306f15ed51b
Reading this record
Response text is shown exactly as the API returned it, with Markdown left unrendered. Line numbers and highlights are added by this site; highlights come from the same patterns the scorer uses, so they show what the verdict rests on. Each model is asked once per day (k = 1): a single change can be sampling noise, which is why every event links the full responses rather than a summary. How scoring works.