VMLU - STEM: leaderboard
VMLU benchmark — accuracy on Vietnamese-language STEM subjects.
Metric: Accuracy (%). Source: vmlu.ai. Saturation forecast: Estimated already saturated. 25 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | QwQ-32B | 81.11 | #421 |
| 2 | GPT-4 | 63.84 | #438 |
| 3 | QwQ 32B-Preview | 63.42 | #651 |
| 4 | Llama 3 70B | 61.7 | #631 |
| 5 | GPT-4o Mini | 58 | #599 |
| 6 | Gemma 2 9B (IT) | 56.31 | #788 |
| 7 | Qwen 2.5 7B Instruct | 55.43 | #862 |
| 8 | ChatGPT | 43.24 | #900 |
| 9 | gemma-7B (IT) | 39.95 | #1496 |
| 10 | Phi-3-small-128k-instruct | 39.31 | #1105 |
| 11 | Phi-3-small-8k-instruct | 38.72 | #967 |
| 12 | Qwen-7B | 30.64 | #1402 |
| 13 | bloom-1b7 | 25.13 | #1628 |
| 14 | bloom-7b1 | 25.08 | #1613 |
| 15 | gemma-2B (IT) | 24.39 | #1559 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vmlu-stem · How It Works · Data refreshed daily, snapshot 2026-10-20.