FIKA-Bench: leaderboard
Metric: Overall strict answer accuracy (%) on FIKA-Bench, open-ended fine-grained recognition questions (311 images from decontaminated public fine-grained datasets and real-life photos, each answer backed by human-verified evidence URLs, kept after strong closed-book models failed); a text-only LLM judge counts only answers at the required granularity as correct, closed-book models (no tools, image and question only); higher is better. Source: arxiv.org. Saturation forecast: Around February 2028. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Kimi K2.6 | 25.1 |
| 2 | Gemini 3.1 Flash Lite | 20.6 |
| 3 | Qwen 3.5 397B A17B | 18.3 |
| 4 | GPT-5 Mini | 17.4 |
| 5 | GLM-5V Turbo | 15.8 |
| 6 | Qwen 3 VL 235B A22B | 14.8 |
| 7 | Qwen 3.5 9B | 9.3 |
Interactive version: theaggregate.ai/benchmark?slug=fika-bench · How It Works · Data refreshed daily, snapshot 2026-10-07.