Benchmarks
357 sources · 1,263 models × 3,079 benchmarks · 162,127 scoresThe 15 released tasks of Synthetic Sciences' FrontierML (seven model, five method and three judgment tasks), with the same 12-hour Lab0 runs, Horizon harness and 0-1 normalization as the 25-task headline: the mean over these tasks of one run's normalized score, clipped to 0-1. Grok 4.7 ran only these tasks. It is a component of the Synthetic Sciences FrontierML factor group. Unrelated to Induction Labs' FrontierML (frontierml).
No signal or difficulty estimate. Too few models have been measured on this benchmark to estimate its difficulty.
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | Claude Fable 5.2 | max · Horizon 0.4.17 (Synthetic Sciences' FrontierML build of mini-swe-agent, under Harbor) | 0.6 | |
| 2 | Claude Opus 5.5 | max · Horizon 0.4.17 (Synthetic Sciences' FrontierML build of mini-swe-agent, under Harbor) | 0.4 | |
| 3 | Gemini 4 Carbon | high · Horizon 0.4.17 (Synthetic Sciences' FrontierML build of mini-swe-agent, under Harbor) | 0.3 | |
| 4 | GPT-6 Astra | max · Horizon 0.4.17 (Synthetic Sciences' FrontierML build of mini-swe-agent, under Harbor) | 0.3 | |
| 5 | Grok 4.7 | x-high · Horizon 0.4.17 (Synthetic Sciences' FrontierML build of mini-swe-agent, under Harbor) | 0.1 |