QIMMA - 3LM MBPP+: leaderboard
TII's 3LM Arabic version of MBPP+: short Python programming problems posed in Arabic, checked against EvalPlus's extended test suites in the QIMMA Arabic leaderboard. pass@1 on a 0-100 scale, higher is better; open-weight models only.
Metric: pass@1 (%). Source: huggingface.co. Saturation forecast: Around February 2027. 83 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 235B A22B 2507 Instruct | 77.51 | #315 |
| 2 | Qwen 3.5 397B A17B | 76.72 | #154 |
| 3 | Qwen 3.5 27B | 74.34 | #192 |
| 4 | Qwen 3.6 35B A3B | 74.07 | #238 |
| 5 | Qwen 3.8 27B | 73.81 | #123 |
| 6 | Qwen 3.6 27B | 73.02 | #182 |
| 7 | Qwen 2.5 72B Instruct | 72.75 | #469 |
| 8 | Qwen 2.5 32B Instruct | 72.75 | #509 |
| 9 | Qwen 3.5 122B A10B | 72.49 | #184 |
| 10 | Gemma 3 27B (IT) | 71.43 | #528 |
| 11 | Llama 3.3 70B Instruct | 71.16 | #541 |
| 12 | Gemma 3 12B (IT) | 70.63 | #672 |
| 13 | Qwen 3.5 35B A3B | 70.63 | #268 |
| 14 | AIC-1 | 69.58 | #235 |
| 15 | Gemma 4 31B (IT) | 68.25 | #178 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=qimma-3lm-mbpp-plus · How It Works · Data refreshed daily, snapshot 2026-11-05.