HallusionBench
HallusionBench - benchmark for language and visual hallucination.
Category
Vision & Language
Max Score
100.0
Score Type
percent
Active
Yes
Model Scores
Scores for HallusionBench
| Model | Score | Date | Verified |
|---|---|---|---|
| Qwen3.6-35B-A3B |
98.02%
|
16.04.2026 | Verified |
| Gemma4-31B-it |
74.26%
|
16.04.2026 | Unverified |
| Qwen3.5-35B-A3B |
79.21%
|
16.04.2026 | Unverified |
| Gemma4-26B-A4B |
61.39%
|
16.04.2026 | Unverified |
| Claude-Sonnet-4.5 |
59.90
|
16.04.2026 | Unverified |
| Qwen3.5-27B |
100.00%
|
16.04.2026 | Unverified |
| Qwen3.5-4B |
50.50%
|
01.02.2026 | Verified |
| Qwen3.5-122B-A10B |
76.24%
|
01.02.2026 | Verified |
| GPT-5-mini 2025-08-07 |
32.67%
|
01.02.2026 | Unverified |
| Qwen3-235B-A22B |
67.33%
|
01.02.2026 | Unverified |
| Qwen3-VL-235B-A22B |
67.33%
|
01.02.2026 | Verified |