reasoning benchmark
Hallusion Bench Leaderboard
Hallusion Bench evaluates image-context reasoning in large visual-language models using 346 images and 1,129 questions.
Updated Sep 5, 2026
Models18
Model coverage18
MetricScore
EvidenceB
Hallusion Bench Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAC | Score70.00% | Percentile100.00% | Participants18 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score69.80% | Percentile94.12% | Participants18 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score67.90% | Percentile88.24% | Participants18 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score67.60% | Percentile82.35% | Participants18 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score67.40% | Percentile76.47% | Participants18 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score66.70% | Percentile70.59% | Participants18 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score66.00% | Percentile64.71% | Participants18 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score65.40% | Percentile58.82% | Participants18 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score64.10% | Percentile52.94% | Participants18 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score63.80% | Percentile47.06% | Participants18 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score63.20% | Percentile41.18% | Participants18 | EvidenceC | Evaluated |
| Rank12 | ModelCO | Score61.50% | Percentile35.29% | Participants18 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score61.50% | Percentile29.41% | Participants18 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score61.10% | Percentile23.53% | Participants18 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score57.60% | Percentile17.65% | Participants18 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score55.16% | Percentile11.76% | Participants18 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score52.90% | Percentile5.88% | Participants18 | EvidenceC | Evaluated |
| Rank18 | ModelLA | Score47.20% | Percentile0.00% | Participants18 | EvidenceC | Evaluated |
Hallusion Bench Highlights
The leading models and scores on this benchmark.
Hallusion Bench Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for Hallusion Bench
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is Hallusion Bench?
What Hallusion Bench measures and how its scores work.
Hallusion Bench is a reasoning benchmark for large visual-language models (LVLMs).
It measures language hallucination and visual illusion through image-context reasoning questions, with results reported as a Score ratio.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- Hallusion Bench
- Modality
- multimodal
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about Hallusion Bench.
Which model scores highest on Hallusion Bench?
Qwen3.5-27B is currently ranked first with 70.00%.
What are the top three models on Hallusion Bench?
The current leaders are Qwen3.5-27B (70.00%), Qwen3.6-35B-A3B (69.80%), and Qwen3.5-35B-A3B (67.90%).
Which Hallusion Bench model has the lowest official input price?
Qwen3.6-35B-A3B has the lowest matched official input price at $0.25 input / $1.5 output per 1M tokens.
Which models are fastest among Hallusion Bench results?
The fastest matched records are Qwen3 VL 4B Thinking (8.77 tok/s via DeepInfra).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.