Onyx Open Source Leaderboard: leaderboard
Open-source model rankings organized into S/A/B/C/D tiers based on aggregate performance across reasoning, coding, math, chat, and instruction following.
Metric: MMLU-Pro. Source: www.onyx.app. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-OSS-120B | 90 | #341 |
| 2 | Qwen 3.5 397B A17B | 87.8 | #146 |
| 3 | DeepSeek V4 Pro | 87.5 | #102 |
| 4 | Nemotron 3 Ultra | 86.8 | #235 |
| 5 | DeepSeek V4 Flash | 86.2 | #137 |
| 6 | Qwen 3.6 27B | 86.2 | #170 |
| 7 | Step 3.5 Flash | 85.8 | #270 |
| 8 | Gemma 4 31B | 85.2 | #200 |
| 9 | Qwen 3.6 35B A3B | 85.2 | #226 |
| 10 | DeepSeek V3.2 | 85 | #205 |
| 11 | MiMo-V2-Flash | 84.9 | #372 |
| 12 | GLM-4.7 | 84.3 | #192 |
| 13 | DeepSeek R1 | 84 | #251 |
| 14 | Nemotron 3 Super | 83.7 | #335 |
| 15 | Llama 4 Maverick | 80.5 | #464 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=onyx-open-source-leaderboard · How It Works · Data refreshed daily, snapshot 2026-10-20.