Cedar - Multimodal - Hindi: leaderboard

Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around 2031. 15 models tracked.

Top models

#ModelScore
1Gemini 2.5 Flash (Thinking)47.73
2Gemma 3 27B (IT)44.38
3Qwen 3 VL 235B A22B (Thinking)41.71
4Claude Sonnet 4.540.82
5GPT-4o (2024-11-20)39.14
6aya-vision-8B33.57
7Mistral Small 3.232.98
8Llama 3.2 11B Instruct32.92
9Pixtral-12B-240929.55
10Qwen 3 VL 8B (Thinking)27.85

Interactive version: theaggregate.ai/benchmark?slug=cedar-multimodal-hindi · How It Works · Data refreshed daily, snapshot 2026-09-25.