Visual Acuity Test (4K): leaderboard
Visual Acuity Test for LLMs — mean character accuracy reading progressively smaller text in a 4K image. Retains each published checkpoint, detail, resolution and reasoning configuration. The source averages over attempted rows, whose counts are retained with each result.
Metric: Mean Accuracy (%). Source: jonathanc.net. Saturation forecast: Rough model projection: around 2026. 47 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (Visual Acuity Run: GPT-5.4%20%28original%20detail%29) | 100 | |
| 2 | GPT-5.5 (Visual Acuity Run: GPT-5.5%20%28none%29) | 100 | |
| 3 | GPT-5.6 (Visual Acuity Run: GPT-5.6%20%28none%29) | 100 | |
| 4 | GPT-5.6-LUNA (Visual Acuity Run: GPT-5.6-LUNA%20%28none%29) | 100 | |
| 5 | GPT-5.6-TERRA (Visual Acuity Run: GPT-5.6-TERRA%20%28none%29) | 100 | |
| 6 | Gemini 2.0 Flash (Visual Acuity Run: Gemini%202.0%20Flash%20%28high%20res%29) | 100 | |
| 7 | Gemini 3 Pro Preview (Visual Acuity Run: Gemini%203%20Pro%20Preview%20%28medium%20res%29) | 97.8 | |
| 8 | GPT-5.6 (Visual Acuity Run: GPT-5.6%20%28none%29%20%28high%20detail%29) | 96.7 | |
| 9 | Gemini 3 Flash Preview (Minimal) (Visual Acuity Run: Gemini%203%20Flash%20Preview%20%28thinking%3A%20minimal%29%20%28high%20res%29) | 95 | |
| 10 | Gemini 3 Pro Preview (Visual Acuity Run: Gemini%203%20Pro%20Preview%20%28high%20res%29) | 95 |
Interactive version: theaggregate.ai/benchmark?slug=visual-acuity-test-4k · How It Works · Data refreshed daily, snapshot 2026-10-20.