MERA Multi - WEIRD: leaderboard
MERA Multi task WEIRD: a public-test Russian image benchmark of image question answering, scored by MERA's task metric (Exact Match or judge score). MERA's 0-1 result scaled to 0-100; higher is better. From the MERA Multi leaderboard (AI Alliance Russia); only submissions MERA shows publicly for the modality are included.
Metric: Score (%). Source: mera.a-ai.ru. Saturation forecast: Around February 2027. 60 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen3 Omni 30B A3B Instruct | 74.1 | #401 |
| 2 | GPT-4.1 | 73.6 | #258 |
| 3 | Qwen 2.5 VL 72B Instruct | 70.3 | #383 |
| 4 | Qwen 3.5 27B | 66.7 | #192 |
| 5 | gemma-4-E4B-it | 66.2 | #577 |
| 6 | Qwen 3.6 27B | 64.8 | #182 |
| 7 | Qwen 3.5 35B A3B | 63.3 | #268 |
| 8 | Gemma 4 26B A4B (IT) | 63 | #293 |
| 9 | Qwen 2.5 VL 32B Instruct | 62.2 | #463 |
| 10 | Qwen 3.6 35B A3B | 61.9 | #238 |
| 11 | Qwen 3.5 122B A10B | 56.8 | #184 |
| 12 | Gemma 4 31B (IT) | 53.3 | #178 |
| 13 | Gemma 4 E2B (IT) | 49.6 | #779 |
| 14 | Qwen 2.5 VL 7B Instruct | 49 | #650 |
| 15 | Qwen2.5-Omni-7B | 45.5 | #633 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mera-multi-weird · How It Works · Data refreshed daily, snapshot 2026-11-05.