MERA - PARus: leaderboard
MERA is the Russian open benchmark run by the AI Alliance Russia; this is its PARus task: plausible alternatives: pick the more likely cause or effect of a Russian sentence. Each MERA task keeps its own native metric, so the boards are not on one shared scale.
Metric: Accuracy (%). Source: mera.a-ai.ru. Saturation forecast: Estimated already saturated. 632 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Median Human | 98.2 | |
| 2 | DeepSeek R1 Distill Qwen 32B | 96.2 | #655 |
| 3 | Qwen 2.5 72B Instruct | 95.8 | #454 |
| 4 | Gemini 1.5 Pro | 95.8 | #445 |
| 5 | Claude 3.7 Sonnet | 95.4 | #249 |
| 6 | Qwen 2 72B | 95.4 | #521 |
| 7 | Gemma 2 27B (IT) | 95 | #645 |
| 8 | Qwen 2 72B Instruct | 94.8 | #576 |
| 9 | Mistral Large 2 (Jul) | 94.8 | #484 |
| 10 | GPT-4o | 94.4 | #343 |
| 11 | Claude Opus 4.5 | 94.4 | #83 |
| 12 | Llama 3.1 70B Instruct | 94 | #561 |
| 13 | Qwen 2.5 72B | 94 | #499 |
| 14 | Qwen 2.5 32B Instruct | 93.6 | #498 |
| 15 | Claude Opus 4.6 | 93.4 | #63 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mera-parus · How It Works · Data refreshed daily, snapshot 2026-10-20.