MERA - MultiQ: leaderboard
MERA is the Russian open benchmark run by the AI Alliance Russia; this is its MultiQ task: multi-hop question answering over Russian text, where the answer needs two documents. Each MERA task keeps its own native metric, so the boards are not on one shared scale.
Metric: F1 (%). Source: mera.a-ai.ru. Saturation forecast: Around January 2027. 632 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Median Human | 92.8 | |
| 2 | Gemma 4 31B (IT) | 67.3 | #176 |
| 3 | GPT-5.2 | 65.85 | #110 |
| 4 | Kimi K2.5 | 65.34 | #144 |
| 5 | GPT-5.4 | 65.22 | #79 |
| 6 | GLM-5.1 | 65.1 | #97 |
| 7 | Gemma 4 26B A4B (IT) | 64.59 | #297 |
| 8 | Mistral Large 2 (Jul) | 63 | #484 |
| 9 | Gemini 2.0 Flash | 62.35 | #342 |
| 10 | Llama 3.1 405B Instruct | 62.34 | #457 |
| 11 | Claude 3.7 Sonnet | 62.26 | #249 |
| 12 | Qwen 3.6 27B | 61.7 | #170 |
| 13 | Llama 3.1 Nemotron 70B Instruct | 61.54 | #554 |
| 14 | Llama 3.1 70B Instruct | 60.67 | #561 |
| 15 | DeepSeek V4 Pro | 60.05 | #102 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mera-multiq · How It Works · Data refreshed daily, snapshot 2026-10-20.