URIAL-Bench - Math: leaderboard
The "Math" subtask of URIAL-Bench. A task from URIAL-Bench, which evaluates base (untuned) language models aligned purely through in-context prompting (Untuned LLMs with Restyled In-context ALignment), measuring instruction-following and helpfulness without fine-tuning.
Metric: Judge Score (0-10). Source: huggingface.co. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4 | 6.8 | #438 |
| 2 | GPT-3.5 Turbo | 6.3 | #865 |
| 3 | DBRX | 5.9 | #1045 |
| 4 | Mixtral 8x7B (v0.1) | 4.85 | #1083 |
| 5 | Yi 34B (Base) | 4.8 | #972 |
| 6 | gemma-7B | 4.35 | #1384 |
| 7 | Phi-2 | 3.8 | #1511 |
| 8 | Llama 2 70B Base | 3.6 | #1025 |
| 9 | Mistral-7B-v0.1 | 3.4 | #1326 |
| 10 | gemma-2B | 3.3 | #1575 |
| 11 | Llama 2 13B Base | 2.85 | #1445 |
| 12 | Yi 6B (Base) | 2.5 | #1472 |
| 13 | Llama 2 7B Base | 1.6 | #1518 |
| 14 | falcon-7B | 1.6 | #1598 |
| 15 | mpt-7B | 1 | #1619 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=urial-bench-math · How It Works · Data refreshed daily, snapshot 2026-10-20.