GPT-5.6-LunaWatching since 11 July 2026
The Meter · Public record
What we watch, every day
We watch models, and we watch the routes a request travels to reach them. Every line gets the same checks, every day, with the same settings, and we record what comes back. We never compare one company's model with another's. A line is only ever compared with its own earlier days, because that is the only comparison that can show it moved.
What we watch
What we watch
Models share one set of checks. Routes have their own.
Models
GPT-5.6 SolWatching since 20 July 2026
Claude Fable 5Watching since 27 July 2026
Claude Opus 5Watching since 27 July 2026
Gemini 3.1 ProWatching since 28 July 2026
Routes
SEE HOW WE MEASURE ↗The fixed parts
The fixed parts
The exam was sealed before the first run. Its fingerprint was anchored to the Bitcoin blockchain on 14 July 2026.
- Sealed instruments
- See the fingerprints ↗
- Dated runs
- When we change our own settings, we close that run and start a new dated one. The earlier days stay published.
- What is sealed
- Items, per-item results, prompts and run schedule. What is sealed and why.
- Published days
- Days publish after validation; a tile counts published days.