Cedar - Multimodal - Spanish: leaderboard
Culturally grounded emotion understanding on Cedar's 400 Spanish image-plus-narrative scenarios: choose which of 14 emotions Spanish-speaking annotators assign, a label chosen because it differs across cultures for the same scene. Accuracy (%) at temperature 0, higher is better (Cedar, Dai et al., Hefei University of Technology, ACL 2026).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around 2028. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Small 3.2 | 48.15 | #1592 |
| 2 | Gemma 3 27B (IT) | 46.65 | #1306 |
| 3 | Claude Sonnet 4.5 | 45.45 | #244 |
| 4 | Qwen 3 VL 235B A22B (Thinking) | 44.53 | #451 |
| 5 | Gemini 2.5 Flash (Thinking) | 44.05 | #452 (Gemini 2.5 Flash) |
| 6 | GPT-4o (2024-11-20) | 41.6 | #864 |
| 7 | Qwen 3 VL 8B (Thinking) | 39.04 | #887 |
| 8 | Llama 3.2 11B Instruct | 28.12 | #5039 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=cedar-multimodal-spanish · How It Works · Data refreshed daily, snapshot 2026-11-12.