FrontierML ($10 wallet)
Agents · 2026-10-13
The same 21 FrontierML end-to-end machine learning tasks, harnesses, review and scoring as the $200 headline, with a $10 wallet per trial for GPUs and OpenRouter tokens: the mean over the 21 tasks of one trial's normalized score (0-1 between measured floor and ceiling anchors). It has the least headroom of the three wallets, and the top two agents are tied within a tenth of a point. It is a fallback component of the FrontierML factor group.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Sonnet 5.5 | 45.5 |
| Claude Opus 5.5 | 45.4 |
| GPT-6 Astra | 37.6 |
| Claude Fable 5.1 | 35.7 |
| GPT-6.1 Sol | 31.6 |
| Gemini 4 Argon | 29.6 |
| Claude Haiku 5.5 | 28.2 |
| Kimi K3.1 Pro | 25.5 |
| Muse Wonder | 21.6 |
| DeepSeek V4.1 Flash 1017 | 21.3 |
| MiniMax M3.1 Flash | 19.2 |
| GLM-5.3 | 18.6 |
| Kimi K3 | 16.2 |
| Muse Spark 1.3 | 14.1 |
| GPT-6 Luna | 13.2 |
| Grok 4.7 | 12.2 |
| Gemini 3.8 Flash | 8.1 |
| Beam 1.0 501B A23B | 7.5 |