Cedar - Multimodal - Japanese: leaderboard
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around 2034. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 235B A22B (Thinking) | 34.34 |
| 2 | Claude Sonnet 4.5 | 32.99 |
| 3 | GPT-4o (2024-11-20) | 32.5 |
| 4 | Gemma 3 27B (IT) | 30.46 |
| 5 | aya-vision-8B | 30.29 |
| 6 | Gemini 2.5 Flash (Thinking) | 29.07 |
| 7 | Mistral Small 3.2 | 27.05 |
| 8 | Qwen 3 VL 8B (Thinking) | 26.39 |
| 9 | Pixtral-12B-2409 | 24.07 |
Interactive version: theaggregate.ai/benchmark?slug=cedar-multimodal-japanese · How It Works · Data refreshed daily, snapshot 2026-09-25.