HAERAE-Vision - Gaming: leaderboard
The 91 gaming questions of HAERAE-Vision (What Users Leave Unsaid, Choi et al., ACL 2026 Findings): real Korean community questions about video games, mostly asked from screenshots in the users' original under-specified wording. A GPT-5-mini judge scores each answer against a 1-5 item checklist (met 1, partly met 0.5, not met 0); the score is the mean checklist credit per question, in percent, averaged over three runs; higher is better.
Metric: Checklist Score (%; original under-specified queries, GPT-5-mini judge, mean of 3 runs). Source: arxiv.org. Saturation forecast: Around February 2027. 45 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 36.67 | #262 |
| 2 | GPT-5 | 32.34 | #163 |
| 3 | GPT-5 Mini | 29.22 | #349 |
| 4 | Grok 4 | 29 | #306 |
| 5 | Gemini 2.5 Flash | 26.7 | #452 |
| 6 | Qwen 3 VL 235B A22B (Thinking) | 23.97 | #451 |
| 7 | Qwen 3 VL 235B A22B Instruct | 23.28 | #521 |
| 8 | Qwen 3 VL 32B (Thinking) | 19.76 | #557 |
| 9 | Qwen 3 VL 30B A3B Instruct | 18.86 | #870 |
| 10 | Qwen 3 VL 32B Instruct | 18.13 | #582 |
| 11 | Qwen 3 VL 30B A3B (Thinking) | 18.06 | #710 |
| 12 | Gemini 2.5 Flash Lite | 17.97 | #958 |
| 13 | Mistral Medium 3.1 | 16.01 | #852 |
| 14 | Gemma 3 27B | 13.75 | #1638 |
| 15 | Qwen 3 VL 8B Instruct | 13.61 | #982 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=haerae-vision-gaming · How It Works · Data refreshed daily, snapshot 2026-11-12.