PRISM-VLM - Instruction Following: leaderboard
The instruction-following axis of PRISM-VLM (NAVER, 2026): each recycled visual question carries one to three verifiable formatting constraints adapted from IFEval, such as length, casing, keywords or output format. Mean share of constraints satisfied, in percent; higher is better.
Metric: Per-axis mean score (%; mean compliance with 1-3 verifiable IFEval-style formatting constraints appended to each item; over PRISM-VLM's 6,238 items recycled from 15 public VLM benchmarks (five seeds of 100 items per benchmark); GPT-5 (low effort) synthesizes the perturbations and grades the open-ended axes; each model at the lowest reasoning effort its provider exposes, temperature 0). Source: arxiv.org. Saturation forecast: Around February 2027. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini (Minimal) | 87.5 | #349 (GPT-5 Mini) |
| 2 | GPT-5.4 Mini | 87.1 | #389 |
| 3 | GPT-5.4 Nano | 86 | #633 |
| 4 | GPT-4o Mini | 84.3 | #1627 |
| 5 | GPT-5 Nano (Minimal) | 83.7 | #888 (GPT-5 Nano) |
| 6 | Gemini 3.1 Flash Lite (Minimal) | 83.6 | #363 (Gemini 3.1 Flash Lite) |
| 7 | Qwen 3 VL 8B Instruct | 82.3 | #982 |
| 8 | GPT-4.1 Nano | 81.7 | #1992 |
| 9 | Gemini 2.5 Flash (Non-reasoning) | 81.4 | #452 (Gemini 2.5 Flash) |
| 10 | Gemini 3 Flash (Minimal) | 81 | #159 (Gemini 3 Flash) |
| 11 | GPT-4.1 Mini | 80.5 | #775 |
| 12 | Molmo2-8B | 79.8 | #1488 |
| 13 | Gemini 2.0 Flash | 79.1 | #702 |
| 14 | Qwen 3 VL 4B Instruct | 78.9 | #1281 |
| 15 | Claude 3 Haiku | 77.8 | #2629 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=prism-vlm-instruction-following · How It Works · Data refreshed daily, snapshot 2026-11-12.