PRISM-VLM - Hallucination Resistance: leaderboard
Metric: Per-axis mean score (%; share of trap questions (absent entity, misleading premise, uncertain image) refused or corrected; 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 March 2028. 42 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 8B Instruct | 77.9 |
| 2 | Gemma 4 E4B | 74.6 |
| 3 | Claude 3 Haiku | 74.5 |
| 4 | Gemini 3 Flash (Minimal) | 73.4 |
| 5 | Qwen 3 VL 4B Instruct | 72.9 |
| 6 | Gemini 2.0 Flash | 71.6 |
| 7 | Gemini 2.0 Flash Lite | 71.6 |
| 8 | Gemini 2.5 Flash Lite | 71 |
| 9 | Gemma 4 E2B | 71 |
| 10 | GPT-4o Mini | 70.8 |
| 11 | Gemini 2.5 Flash (Non-reasoning) | 70.6 |
| 12 | Qwen 3.5 4B (Non-reasoning) | 69.9 |
| 13 | Qwen 3.5 9B (Non-reasoning) | 68 |
| 14 | Claude Haiku 4.5 | 67.5 |
| 15 | Nova 2 Lite | 67.1 |
Interactive version: theaggregate.ai/benchmark?slug=prism-vlm-hallucination-resistance · How It Works · Data refreshed daily, snapshot 2026-09-26.