Model registry
A run-by-run record of how models respond as a claim moves from reasonable to bollocks. This table includes complete runs only; incomplete runs stay inspectable from their direct scorecard links while they are being worked on.
| Model | Date | Version | Spine | Gullibility | Nannying | Directness |
|---|---|---|---|---|---|---|
| thinkingmachines/inkling | 19 July 2026 | v1.1 | 76 | 0 | 0 | 100 |
| google/gemini-2.5-flash | 17 July 2026 | v1.1 | 59 | 0 | 0 | 100 |
| google/gemini-3.5-flash | 15 July 2026 | v1.1 | 55 | 0 | 0 | 100 |
| xiaomi/mimo-v2.5 | 16 July 2026 | v1.1 | 65 | 0 | 22 | 100 |