Arcee MMLU-Pro (BenchLM): leaderboard
Arcee’s launch-chart comparison on professional academic multiple-choice questions. This source-specific snapshot is kept distinct from standardized MMLU-Pro evaluations. These are published results, not BenchLM ability estimates.
Metric: Accuracy (%). Source: benchlm.ai. Saturation forecast: Estimated already saturated. 74 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.7 Max | 89.6 | #85 |
| 2 | Claude Opus 4.5 | 89.5 | #87 |
| 3 | Qwen 3.6 Plus | 88.5 | #124 |
| 4 | Qwen 3.7 Plus | 88.5 | #98 |
| 5 | Qwen 3.5 397B A17B | 87.8 | #154 |
| 6 | DeepSeek V4 Pro (0813) | 87.5 | #58 |
| 7 | Kimi K2.5 | 87.1 | #152 |
| 8 | Kimi K2.5 (Thinking) | 87.1 | #152 (Kimi K2.5) |
| 9 | Nemotron 3 Ultra | 86.8 | #245 |
| 10 | Qwen 3.5 122B A10B | 86.7 | #184 |
| 11 | Solar Pro 4 | 86.3 | #230 |
| 12 | Qwen 3.6 27B | 86.2 | #182 |
| 13 | DeepSeek V4 Flash (0731) | 86.2 | #84 |
| 14 | Qwen 3.5 27B | 86.1 | #192 |
| 15 | GLM-5 | 85.7 | #149 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=arcee-mmlu-pro-benchlm · How It Works · Data refreshed daily, snapshot 2026-11-05.