Benchmark
Hallusion Bench
What is Hallusion Bench?
A comprehensive benchmark designed to evaluate image-context reasoning in large visual-language models (LVLMs) by challenging models with 346 images and 1,129…
- Released
- 2023-10-23
- Modality
- multimodal
- Evaluates
- reasoning, vision
- Openness
- unknown
- Importer
- LLM Stats
- Reported scores
- 18
Source provenance
Hallusion Bench results and reported scores
LLM Stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| LFM2.5-VL-3B | Liquid AI | 0.472 | 2026-08-12 | evidence |
| North Micro Vision Instruct | Cohere | 0.615 | 2026-08-12 | evidence |
| Qwen2.5 VL 72B Instruct | Qwen | 0.5516 | 2025-01-26 | evidence |
| Qwen2.5 VL 7B Instruct | Qwen | 0.529 | 2025-01-26 | evidence |
| Qwen3 VL 235B A22B Instruct | Qwen | 0.632 | 2025-09-22 | evidence |
| Qwen3 VL 235B A22B Thinking | Qwen | 0.667 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Instruct | Qwen | 0.615 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Thinking | Qwen | 0.66 | 2025-09-22 | evidence |
| Qwen3 VL 32B Instruct | Qwen | 0.638 | 2025-09-22 | evidence |
| Qwen3 VL 32B Thinking | Qwen | 0.674 | 2025-09-22 | evidence |
| Qwen3 VL 4B Instruct | Qwen | 0.576 | 2025-09-22 | evidence |
| Qwen3 VL 4B Thinking | Qwen | 0.641 | 2025-09-22 | evidence |
| Qwen3 VL 8B Instruct | Qwen | 0.611 | 2025-09-22 | evidence |
| Qwen3 VL 8B Thinking | Qwen | 0.654 | 2025-09-22 | evidence |
| Qwen3.5-122B-A10B | Qwen | 0.676 | 2026-02-24 | evidence |
| Qwen3.5-27B | Qwen | 0.7 | 2026-02-24 | evidence |
| Qwen3.5-35B-A3B | Qwen | 0.679 | 2026-02-24 | evidence |
| Qwen3.6-35B-A3B | Qwen | 0.698 | 2026-04-16 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.