MERA - ruMultiAr: leaderboard
MERA is the Russian open benchmark run by the AI Alliance Russia; this is its ruMultiAr task: multi-step arithmetic in Russian. Each MERA task keeps its own native metric, so the boards are not on one shared scale.
Metric: EM (%). Source: mera.a-ai.ru. Saturation forecast: Estimated already saturated. 632 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 100 | #79 |
| 2 | GPT-OSS-20B | 100 | #518 |
| 3 | Qwen 3.5 27B | 100 | #187 |
| 4 | Qwen 3 Next 80B A3B (Thinking) | 100 | #320 (Qwen 3 Next 80B A3B) |
| 5 | Qwen 3 30B A3B 2507 (Thinking) | 100 | #376 (Qwen 3 30B A3B 2507) |
| 6 | GLM-5.1 | 99.9 | #97 |
| 7 | Qwen 3 235B A22B 2507 (Thinking) | 99.9 | #260 (Qwen 3 235B A22B 2507) |
| 8 | Claude Opus 4.6 | 99.8 | #63 |
| 9 | Qwen 3.5 122B A10B | 99.8 | #173 |
| 10 | Qwen 3 4B 2507 (Thinking) | 99.71 | #538 (Qwen 3 4B 2507) |
| 11 | Qwen 3 14B | 99.22 | #537 |
| 12 | Qwen 3.6 27B | 99.12 | #170 |
| 13 | Qwen 3.6 35B A3B | 98.83 | #226 |
| 14 | Qwen 3 8B | 98.54 | #674 |
| 15 | Qwen 3 4B | 98.54 | #829 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mera-rumultiar · How It Works · Data refreshed daily, snapshot 2026-10-20.