HELM VHELM - Flickr30K: leaderboard
The "Flickr30K" subtask of HELM VHELM. Stanford HELM's VHELM project, the holistic evaluation of vision-language models: image-grounded question answering, OCR, chart and diagram reading, spatial reasoning, robustness, bias, fairness and toxicity scenarios scored under one ha...
Metric: Prometheus Vision rating. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20241022) | 3.22 | #291 |
| 2 | GPT-4 Turbo | 3.08 | #406 |
| 3 | Claude 3.5 Sonnet (20240620) | 3.07 | #298 |
| 4 | O4 Mini (2025-04-16) | 3.01 | #181 |
| 5 | GPT-4o (2024-11-20) | 2.97 | #378 |
| 6 | GPT-4.1 Mini | 2.96 | #351 |
| 7 | GPT-4o (2024-05-13) | 2.96 | #326 |
| 8 | GPT-4.1 (2025-04-14) | 2.95 | #250 |
| 9 | GPT-4o (2024-08-06) | 2.95 | #335 |
| 10 | O1 (2024-12-17) | 2.95 | #152 |
| 11 | GPT-4o Mini (2024-07-18) | 2.94 | #561 |
| 12 | O3 (2025-04-16) | 2.93 | #123 |
| 13 | GPT-4.5 | 2.88 | #276 |
| 14 | Gemini 2.0 Flash | 2.81 | #344 |
| 15 | GPT-4.1 Nano | 2.78 | #735 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-vhelm-flickr30k · How It Works · Data refreshed daily, snapshot 2026-10-23.