VideoJev Evaluation - LongVideoBench validation without subtitles: leaderboard
VideoJev Evaluation - LongVideoBench validation without subtitles. Table2 source protocol; VideoJev Joint inference; Instruct and Thinking checkpoints/runs preserved. Molmo2 32K context and Nemotron128K context. Missing source cells retained as absent. Metric: Accuracy (%). Primary published table; source configurations remain distinct.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around January 2027. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seed 2.0 Pro | 80.3 | #114 |
| 2 | Qwen 3.6 35B A3B (Thinking) | 74.9 | #217 (Qwen 3.6 35B A3B) |
| 3 | Qwen 3.5 35B A3B (Thinking) | 74.5 | #250 (Qwen 3.5 35B A3B) |
| 4 | Qwen 3.5 4B (Thinking) | 73.6 | #463 (Qwen 3.5 4B) |
| 5 | Qwen 3 VL 30B A3B (Thinking) | 69.1 | #341 (Qwen 3 VL 30B A3B) |
| 6 | Qwen 3 VL 4B (Thinking) | 63.4 | #469 (Qwen 3 VL 4B) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=videojev-evaluation-longvideobench-validation-without-subtitles · How It Works · Data refreshed daily, snapshot 2026-10-16.