long context benchmark
LVBench Leaderboard
LVBench is an extreme long video understanding benchmark designed to evaluate multimodal models on videos up to two hours in duration. It contains 6 major categories and 21 subcategories, with videos averaging five times longer than existing datasets. The benchmark addresses applications requiring comprehension of extremely long videos.
Updated Aug 17, 2026
Models25
Model coverage25
MetricScore
EvidenceB
LVBench Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelGO | Score85.4% | Percentile100.0% | Participants25 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score81.8% | Percentile95.8% | Participants25 | EvidenceC | Evaluated |
| Rank03 | ModelBY | Score78.0% | Percentile91.7% | Participants25 | EvidenceC | Evaluated |
| Rank04 | ModelBY | Score76.8% | Percentile87.5% | Participants25 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score76.2% | Percentile83.3% | Participants25 | EvidenceC | Evaluated |
| Rank06 | ModelMA | Score75.9% | Percentile79.2% | Participants25 | EvidenceC | Evaluated |
| Rank07 | ModelAC | Score74.4% | Percentile75.0% | Participants25 | EvidenceC | Evaluated |
| Rank08 | ModelAC | Score73.6% | Percentile70.8% | Participants25 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score71.4% | Percentile66.7% | Participants25 | EvidenceC | Evaluated |
| Rank10 | ModelAC | Score71.4% | Percentile62.5% | Participants25 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score67.7% | Percentile58.3% | Participants25 | EvidenceC | Evaluated |
| Rank12 | ModelAC | Score63.8% | Percentile54.2% | Participants25 | EvidenceC | Evaluated |
| Rank13 | ModelAC | Score63.6% | Percentile50.0% | Participants25 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score62.6% | Percentile45.8% | Participants25 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score62.5% | Percentile41.7% | Participants25 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score59.2% | Percentile37.5% | Participants25 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score58.0% | Percentile33.3% | Participants25 | EvidenceC | Evaluated |
| Rank18 | ModelAC | Score56.2% | Percentile29.2% | Participants25 | EvidenceC | Evaluated |
| Rank19 | ModelAC | Score55.8% | Percentile25.0% | Participants25 | EvidenceC | Evaluated |
| Rank20 | ModelAC | Score53.5% | Percentile20.8% | Participants25 | EvidenceC | Evaluated |
| Rank21 | ModelAC | Score49.0% | Percentile16.7% | Participants25 | EvidenceC | Evaluated |
| Rank22 | ModelAC | Score47.3% | Percentile12.5% | Participants25 | EvidenceC | Evaluated |
| Rank23 | ModelAC | Score45.3% | Percentile8.3% | Participants25 | EvidenceC | Evaluated |
| Rank24 | ModelAM Nova ProAmazon | Score41.6% | Percentile4.2% | Participants25 | EvidenceC | Evaluated |
| Rank25 | ModelAMNova LiteAmazon | Score40.4% | Percentile0.0% | Participants25 | EvidenceC | Evaluated |
LVBench Highlights
The leading models and scores on this benchmark.
LVBench Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for LVBench
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis lvbench 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.
- 01GOGemini 3.7 FlashGoogle
- Score
- 85.4%
- Price
- $0.75 input / $3.8 output per 1M tokens
- Speed
- Up to 13 tok/s via Google
Strengths
- Ranks #1 of 25 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures LVBench, not total model capability
- 02ACQwen3.8 MaxAlibaba Cloud / Qwen Team
- Score
- 81.8%
- Price
- $2.0 input / $6.0 output per 1M tokens
Strengths
- Ranks #2 of 25 compared models
- 96th percentile on this benchmark
- C evidence result
Considerations
- This result measures LVBench, not total model capability
- 03BYSeed 2.1 ProByteDance
- Score
- 78.0%
Strengths
- Ranks #3 of 25 compared models
- 92th percentile on this benchmark
- C evidence result
Considerations
- This result measures LVBench, not total model capability
- 04BYSeed 2.1 TurboByteDance
- Score
- 76.8%
Strengths
- Ranks #4 of 25 compared models
- 88th percentile on this benchmark
- C evidence result
Considerations
- This result measures LVBench, not total model capability
- 05ACQwen3.7-PlusAlibaba Cloud / Qwen Team
- Score
- 76.2%
- Price
- $0.50 input / $3.0 output per 1M tokens
Strengths
- Ranks #5 of 25 compared models
- 83th percentile on this benchmark
- C evidence result
Considerations
- This result measures LVBench, not total model capability
Selection summary
Best AI Models for LVBench
Gemini 3.7 Flash currently leads LVBench with 85.4%. 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 LVBench?
What LVBench measures and how its scores work.
LVBench is an extreme long video understanding benchmark designed to evaluate multimodal models on videos up to two hours in duration. It contains 6 major categories and 21 subcategories, with videos averaging five times longer than existing datasets. The benchmark addresses applications requiring comprehension of extremely long videos.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- LVBench
- Modality
- multimodal
- Primary category
- long context
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- lvbench|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about LVBench.
Which model scores highest on LVBench?
Gemini 3.7 Flash is currently ranked first with 85.4%.
What does LVBench measure?
LVBench is an extreme long video understanding benchmark designed to evaluate multimodal models on videos up to two hours in duration. It contains 6 major categories and 21 subcategories, with videos averaging five times longer than existing datasets. The benchmark addresses applications requiring comprehension of extremely long videos.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
25 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.