Benchmarks
357 sources · 1,263 models × 3,079 benchmarks · 162,127 scoresSurge's GDP.pdf board as rescored between its November 6 and November 18, 2026 site builds, on the 100 held-out professional PDF tasks. The page has no changelog and names no metric, judge or run count. Surge's GitHub harness, unchanged since August 14, still describes five no-tools Inspect runs per task that pass only when Gemini 3.5 Flash judges every rubric criterion met, and the launch post still says no frontier model cleared 25%. Every published value is a multiple of 0.2, which fits five runs on each of the 100 tasks. Kept apart from the original board because every run it kept rose, by 16.2 to 26.0 points (Gemini 4 Argon High, 28.4 to 53.6; Grok 4.7 xHigh, 22.8 to 48.8).
413th strongest of 818 fitted benchmarks
Signal-to-noise +5.85
148th hardest of 818 fitted benchmarks
Capability needed 167.9
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | Gemini 4 Carbon | high | 56.4 | |
| 2 | Muse Wonder | max | 54.8 | |
| 3 | Gemini 4 Hydrogen | x-high | 54.2 | |
| 4 | Gemini 4 Argon | high | 53.6 | |
| 5 | Gemini 4 Neon | high | 53.4 | |
| 6 | Muse Spark 1.3 | max | 53.0 | |
| 7 | Claude Fable 5.2 | max | 51.8 | |
| 8 | GPT-6.1 Sol | max | 51.6 | |
| 9 | Claude Sonnet 5.5 | max | 51.2 | |
| 10 | GPT-6 Astra | max | 50.4 | |
| 11 | Claude Fable 5.1 | max | 49.4 | |
| 12 | GPT-6.2 Astra | max | 49.2 | |
| 13 | Muse Ember | max | 49.2 | |
| 14 | Claude Opus 5.5 | max | 48.8 | |
| 15 | GPT-6.2 Sol | max | 48.8 | |
| 16 | Grok 4.7 | x-high | 48.8 | |
| 17 | Kimi K3.1 Pro | max | 46.6 | |
| 18 | Gemini 3.8 Flash | high | 44.2 | |
| 19 | Hy4 | high | 43.6 | |
| 20 | Inkling 2 Small | 0.99 | 43.6 | |
| 21 | Qwen4-Max | x-high | 43.6 | |
| 22 | Kimi K3 | max | 39.8 | |
| 23 | GLM-5.3 | max | 36.6 | |
| 24 | MiniMax M3.1 Flash | max | 27.0 |
Benchmarks outside its family whose model rankings track it most reliably (Spearman ρ over at least 10 shared models).