LVBench
We show this table for reference; we do not rank on it.
A long-video understanding benchmark for retrieving and reasoning over information distributed across extended video inputs.
Benchmark score on LVBench — November 13, 2026
We compile the LVBench rows from provider self-reports. Gemini 4 Argon leads the table at 91.7%, followed by Gemini 3.8 Flash (87.1%) and Gemini 3.7 Flash (85.4%). We do not use these results to rank models overall.
Gemini 4 Argon
Gemini 3.8 Flash
Gemini 3.7 Flash
7 modelsMultimodal & GroundedCurrentDisplay onlyUpdated November 13, 2026
Benchmark score table (7 models)
All 7 results on this page were reported by the labs that built the models.
| Rank | Model / configuration | Score | Who ran it | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Gemini 4 ArgonGoogleReported by Google | 91.7% | Reported by Google | Not reported | Closed |
| 2 | Gemini 3.8 FlashGoogleReported by Google | 87.1% | Reported by Google | Not reported | Closed |
| 3 | Gemini 3.7 FlashGoogleReported by Google | 85.4% | Reported by Google | Not reported | Closed |
| 4 | Qwen3.8 MaxAlibabaReported by Alibaba | 81.8% | Reported by Alibaba | Not reported | Open |
| 5 | Qwen3.8-Omni-FlashAlibabaReported by Alibaba | 76.9% | Reported by Alibaba | Not reported | Closed |
| 6 | Qwen3.8-Flash-NextAlibabaReported by Alibaba | 76.6% | Reported by Alibaba | Not reported | Open |
| 7 | LFM3-VL-2BLiquidAIReported by LiquidAI | 41.2% | Reported by LiquidAI | Not reported | Open |
- Lab-reported
- Benchmark-run
- Independent
Among the reported LVBench rows, Gemini 4 Argon is first at 91.7%. The third row is 6.3 points behind. The broader top-10 range is 50.5 points, so the table still separates the published systems.
7 models have been evaluated on LVBench. The benchmark falls in the Multimodal & Grounded category. LVBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About LVBench
Year
2026
Tasks
Long-form video question answering
Format
Long-video understanding score
Difficulty
Extended temporal reasoning
We store Qwen's standard LVBench result without the separate memory system. The provider also reports an LVBench-with-memory row, which remains distinct and is not collapsed into this value.
Freshness and provenance
Version
LVBench 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does LVBench measure?
A long-video understanding benchmark for retrieving and reasoning over information distributed across extended video inputs.
Which model scores highest on LVBench?
Gemini 4 Argon by Google currently leads with a score of 91.7% on LVBench.
How many models are evaluated on LVBench?
7 AI models have published results on LVBench in the BenchLM catalog.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.