EpiQAL - Multi-Step Inference: leaderboard
Metric: Exact Match (%; zero-shot, no CoT, full article). Source: arxiv.org. Saturation forecast: Around December 2026. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Mistral 7B Instruct (v0.3) | 78.9 |
| 2 | Qwen 3 32B | 74.3 |
| 3 | GPT-4.1 Nano | 67.8 |
| 4 | Mistral Large 2 (Nov) Instruct (2411) | 64.4 |
| 5 | Qwen 3 30B A3B 2507 Instruct | 57.3 |
| 6 | GPT-5 Mini | 53.3 |
| 7 | Qwen 3 8B | 50.8 |
| 8 | Llama 3.1 8B Instruct | 31.8 |
| 9 | Phi-4 Mini Instruct | 24.9 |
| 10 | GPT-4o Mini | 22.2 |
| 11 | Llama 3.2 3B Instruct | 15.7 |
Interactive version: theaggregate.ai/benchmark?slug=epiqal-multi-step-inference · How It Works · Data refreshed daily, snapshot 2026-09-25.