Benchmarks / Reasoning
NanoGPT Speedrun
Uses Prime Intellect's source-reported share of the human record gap closed for each model's best validated optimizer-speedrun result. Score: share of the human record gap closed. Collected by Jake Boggs' benchmark dashboard.
NanoGPT Speedrun
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Fable 5 | max | 81.7% |
| 2 | Opus 5 | max | 53.6% |
| 3 | Kimi K3 | max | 52.2% |
| 4 | Opus 4.8 | max | 39.4% |
| 5 | GPT-5.6 Sol | max | 35.9% |
| 6 | GPT-5.6 Sol Pro | max | 33.6% |
| 7 | Sonnet 5 | max | 26.8% |
| 8 | GPT-5.6 Luna | max | 26.1% |
| 9 | Grok 4.5 | max | 24.6% |
| 10 | Qwen 3.8 Max | max | 24.6% |
| 11 | GLM 5.2 | max | 20.3% |
| 12 | DeepSeek V4 Pro | max | 12.3% |
| 13 | GPT-5.6 Terra | max | 11.0% |
| 14 | Grok 4.6 | max | 10.1% |
| 15 | Muse Spark 1.2 | max | 8.7% |
| 16 | Muse Spark 1.1 | max | 8.4% |
| 17 | GPT-5.5 | xhigh | 8.1% |
| 18 | Kimi K2.7 | none | 7.2% |