Hedge-Bench Forecast Accuracy
Finance & Accounting · 2026-06-02
Share of Hedge-Bench's Yes/No market forecasts that match the resolved outcome, over 92 environments and three trials. The majority-class baseline is 53.26%.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 73.2 |
| Claude Fable 5.1 | 71.4 |
| GPT-5.6 Sol | 69.6 |
| Gemini 3.8 Flash | 67.0 |
| GLM-5.3 | 67.0 |
| Kimi K3 | 67.0 |
| GPT-5.6 Terra | 64.5 |
| Grok 4.6 | 64.5 |
| DeepSeek V4 Pro 0813 | 62.0 |
| Muse Spark 1.3 | 61.2 |