reasoning benchmark
Hallusion Bench Leaderboard
A comprehensive benchmark designed to evaluate image-context reasoning in large visual-language models (LVLMs) by challenging models with 346 images and 1,129 carefully crafted questions to assess language hallucination and visual illusion
Updated Aug 17, 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 | Score70.0% | Percentile100.0% | Participants18 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score69.8% | Percentile94.1% | Participants18 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score67.9% | Percentile88.2% | Participants18 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score67.6% | Percentile82.3% | Participants18 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score67.4% | Percentile76.5% | Participants18 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score66.7% | Percentile70.6% | Participants18 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score66.0% | Percentile64.7% | Participants18 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score65.4% | Percentile58.8% | Participants18 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score64.1% | Percentile52.9% | Participants18 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score63.8% | Percentile47.1% | Participants18 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score63.2% | Percentile41.2% | Participants18 | EvidenceC | Evaluated |
| Rank12 | ModelCO | Score61.5% | Percentile35.3% | Participants18 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score61.5% | Percentile29.4% | Participants18 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score61.1% | Percentile23.5% | Participants18 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score57.6% | Percentile17.6% | Participants18 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score55.2% | Percentile11.8% | Participants18 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score52.9% | Percentile5.9% | Participants18 | EvidenceC | Evaluated |
| Rank18 | ModelLA | Score47.2% | Percentile0.0% | 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.
Ranking basisThis hallusion bench AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01ACQwen3.5-27BAlibaba Cloud / Qwen Team
- Score
- 70.0%
- Price
- $0.30 input / $2.4 output per 1M tokens
- Speed
- Up to 6.7 tok/s via Novita
Strengths
- Ranks #1 of 18 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures Hallusion Bench, not total model capability
- 02ACQwen3.6-35B-A3BAlibaba Cloud / Qwen Team
- Score
- 69.8%
- Price
- $0.25 input / $1.5 output per 1M tokens
Strengths
- Ranks #2 of 18 compared models
- 94th percentile on this benchmark
- C evidence result
Considerations
- This result measures Hallusion Bench, not total model capability
- 03ACQwen3.5-35B-A3BAlibaba Cloud / Qwen Team
- Score
- 67.9%
- Price
- $0.25 input / $2.0 output per 1M tokens
Strengths
- Ranks #3 of 18 compared models
- 88th percentile on this benchmark
- C evidence result
Considerations
- This result measures Hallusion Bench, not total model capability
- 04ACQwen3.5-122B-A10BAlibaba Cloud / Qwen Team
- Score
- 67.6%
- Price
- $0.40 input / $3.2 output per 1M tokens
Strengths
- Ranks #4 of 18 compared models
- 82th percentile on this benchmark
- C evidence result
Considerations
- This result measures Hallusion Bench, not total model capability
- 05ACQwen3 VL 32B ThinkingAlibaba Cloud / Qwen Team
- Score
- 67.4%
Strengths
- Ranks #5 of 18 compared models
- 76th percentile on this benchmark
- C evidence result
Considerations
- This result measures Hallusion Bench, not total model capability
Selection summary
Best AI Models for Hallusion Bench
Qwen3.5-27B currently leads Hallusion Bench with 70.0%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is Hallusion Bench?
What Hallusion Bench measures and how its scores work.
A comprehensive benchmark designed to evaluate image-context reasoning in large visual-language models (LVLMs) by challenging models with 346 images and 1,129 carefully crafted questions to assess language hallucination and visual illusion
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
- Yes
- Evaluation key
- hallusion-bench|llm-stats-current
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 70.0%.
What does Hallusion Bench measure?
A comprehensive benchmark designed to evaluate image-context reasoning in large visual-language models (LVLMs) by challenging models with 346 images and 1,129 carefully crafted questions to assess language hallucination and visual illusion
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
18 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.