Evaluation sources Official evaluation
MirrorCode
Epoch AI / Coding · Long-running software reproduction
Across NexGateHub AI News Reference only
Evidence budgetNot scored
Upstream data time09/23 00:42
Last successful sync10/18 14:05
What it measures and how
Published data covers few models with long runtimes. Samples and run conditions do not support a category consensus.
How this evidence is used
Reference only; excluded from overall and category scores.
Evaluation results
Fixed rules select one representative configuration per public model. Anonymous test models are omitted. Original ranks are retained, with up to 30 models per evaluation.
| Source rank | Source model name | Raw result | Representative configuration |
|---|---|---|---|
| 1 | claude-opus-5-5_maxAnthropic | 77.4% | Max reasoning |
| 2 | claude-fable-5-1_highAnthropic | 73.3% | High reasoning |
| 3 | claude-fable-5_highAnthropic | 63.9% | High reasoning |
| 4 | gpt-6-astra_highOpenAI | 46.7% | High reasoning |
| 5 | claude-opus-4-7_highAnthropic | 31.1% | High reasoning |
| 6 | gpt-5.6-sol_highOpenAI | 20.0% | High reasoning |
| 7 | gpt-5.4-2026-03-05_highOpenAI | 15.6% | High reasoning |
| 8 | gpt-5.5_highOpenAI | 10.0% | High reasoning |
| 9 | gemini-3.1-pro-preview_highGoogle | 8.9% | High reasoning |
Evaluation limits & attribution
Reference only; excluded from overall and category scores.
Data license: CC BY 4.0 · Epoch AI's own evaluation data.
Results are published by Epoch AI. Original scores and the NexGateHub consensus index use different scales and cannot be added directly.