83.3%
best 94.1% · Gemini 3.1 Pro Pr…
Not separable from Kimi K2.7 Code at the measured 1.78 point band. The order is printed; the gap is not claimed.
Each rail is one classification. Every ranked model measured on it is a tick; this model is the orange marker. A hatched rail means this model has no measurement in that classification and its weight was redistributed.
One card per benchmark this model has been measured on. The rail shows where it sits against every other ranked model measured on the same benchmark. Where the evaluator keys on a harness, the harness is named, because the same model scores differently under a different one.
Every number on this page, with who measured it, the interval they published, the harness it was run under, and where to go and read it. Nothing here was measured by Ferrox Labs.
| Benchmark | Class | Published | Interval (normalised) | Measured by | Licence | Harness / effort | |
|---|---|---|---|---|---|---|---|
| APEX Agents | agents | 23.7% | 17.6 to 29.8 | APEX | CC-BY-4.0 | / high | source |
| LMArena Agent | agents | 0.017 | 18.3 to 26.7 | LMArena | CC-BY-4.0 | source | |
| OSWorld | agents | 72.1% | not published | OSWorld (XLANG Lab) | none stated | General model / 100 steps | source |
| Terminal-Bench 2 | agents | 53.4% | 47.9 to 58.9 | tbench.ai | CC-BY-4.0 | Simplai Agent | source |
| ALE-Bench | coding | 1327 | not published | ALE-Bench (Sakana AI with AtCoder) | CC-BY-4.0 | / medium | source |
| DeepSWE | coding | 29.9% | 25.8 to 34.0 | deepswe.datacurve.ai | CC-BY-4.0 | mini-swe-agent / high | source |
| FrontierCode | coding | 24.3% | not published | cognition.com | CC-BY-4.0 | claude-code / max | source |
| SWE-bench Verified (Epoch's own run) | coding | 75.2% | 71.4 to 79.1 | Epoch AI | CC-BY-4.0 | source | |
| LMArena Text (style-controlled) | preference | 1472 | 81.2 to 82.3 | LMArena | CC-BY-4.0 | source | |
| LMArena WebDev | preference | 1522 | 64.6 to 66.0 | LMArena | CC-BY-4.0 | source | |
| ARC-AGI-2 | reasoning | 58.3% | not published | ARC Prize | CC-BY-4.0 | / max | source |
| FrontierMath | reasoning | 32.4% | 26.9 to 37.9 | Epoch AI | CC-BY-4.0 | / 16k | source |
| GPQA Diamond (Epoch's own run) | reasoning | 83.3% | 78.1 to 88.5 | Epoch AI | CC-BY-4.0 | / high | source |
| OTIS Mock AIME 2024-2025 | reasoning | 75.6% | 62.9 to 88.3 | Epoch AI | CC-BY-4.0 | / high | source |
Configurations rolled up: 10. Rule: median observed configuration per benchmark (lower median, always a real measurement). Harnesses seen: claude-code, general-model, mini-swe-agent, simplai-agent.
Snapshot 2026-08-30-62e4e85f043b, manifest dc46131315046f1e. Grades are positional: position in the measured field, as a percentile of rank among ranked entries, n=37.