Benchmark Almanac

ODV-Bench

StreamForest (FT-drive) 7B 71.2% tops its leaderboard Qwen2.5-VL-7B 55.6% leads the models we track
Seed2.0 Pro 72.5% leads since February 2026
Fading

Streaming video models answer real-time questions about autonomous driving footage.

Introduced with the StreamForest model, a NeurIPS 2025 spotlight. The format field is a guess.

Released
September 2025
Kind
Video understanding
Level
Everyday
Scored
Checked automatically
Answers
Multiple choice
Last in a launch post
April 2026, dropped before it saturated
Last model added
April 2026
Scores refreshed
November 12th, 2026
11
Entries on its own leaderboard, 5 models we track
1
Launch post lists it, from 1 lab
1
System card reports it, from 1 lab

StreamForest (FT-drive) 7B tops the leaderboard with 71.2%

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Qwen2.5-VL-7B, at 55.6%.

Seed2.0 Pro leads with 72.5%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

The newest model with a score here came out in January 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 71.2 StreamForest (FT-drive) 7B
  2. 59.9 StreamForest 7B
  3. 55.6 Qwen2.5-VL-7B
  4. 54.5 VideoChat-Online 4B
  5. 54.4 VideoChat-Flash 7B
  6. 54.2 InternVL2.5-8B
  7. 51.6 LLaVA-OneVision 7B
  8. 50.2 LongVA-7B
  9. 49.8 MiniCPM-V-2_6
  10. 45.2 Dispider 7B
  11. 35.7 Flash-Vstream 7B
  1. 1 StreamForest (FT-drive) 7B frames=1fps 71.2%
  2. 2 StreamForest 7B frames=1fps 59.9%
  3. 3 Qwen2.5-VL-7B frames=1fps 55.6%
  4. 4 VideoChat-Online 4B frames=1fps 54.5%
  5. 5 VideoChat-Flash 7B frames=256 54.4%
  6. 6 InternVL2.5-8B frames=32 54.2%
  7. 7 LLaVA-OneVision 7B frames=64 51.6%
  8. 8 LongVA-7B frames=64 50.2%
  9. 9 MiniCPM-V-2_6 frames=64 49.8%
  10. 10 Dispider 7B frames=1fps 45.2%
  11. 11 Flash-Vstream 7B frames=1fps 35.7%

Source: Official ODV-Bench leaderboard

Sources: Labs’ model cards and official ODV-Bench leaderboard

#EntryScoreSettingDateSource
1 StreamForest (FT-drive) 7B 71.2% frames=1fps Official leaderboard
2 StreamForest 7B 59.9% frames=1fps Official leaderboard
3 Qwen2.5-VL-7B 55.6% frames=1fps Jan 2025 Official leaderboard
4 VideoChat-Online 4B 54.5% frames=1fps Official leaderboard
5 VideoChat-Flash 7B 54.4% frames=256 Official leaderboard
6 InternVL2.5-8B 54.2% frames=32 Dec 2024 Official leaderboard
7 LLaVA-OneVision 7B 51.6% frames=64 Aug 2024 Official leaderboard
8 LongVA-7B 50.2% frames=64 Jun 2024 Official leaderboard
9 MiniCPM-V-2_6 49.8% frames=64 Aug 2024 Official leaderboard
10 Dispider 7B 45.2% frames=1fps Official leaderboard
11 Flash-Vstream 7B 35.7% frames=1fps Official leaderboard
#ModelScoreSettingDateSource
1 Seed2.0 Pro 72.5% Feb 2026 Lab model card
2 Seed2.0 Lite (0428) 66% Apr 2026 Lab model card
3 Qwen2.5-VL-7B 55.6% frames=1fps Jan 2025 Official leaderboard
4 InternVL2.5-8B 54.2% frames=32 Dec 2024 Official leaderboard
5 LLaVA-OneVision 7B 51.6% frames=64 Aug 2024 Official leaderboard
6 LongVA-7B 50.2% frames=64 Jun 2024 Official leaderboard
7 MiniCPM-V-2_6 49.8% frames=64 Aug 2024 Official leaderboard

Labs report scores for 2 models in their cards. Switch the scores above to rank them together.

Number one over time 2 models have held number one with claims included since its release in September 2025

Lab claims count from the day the card came out, independent runs from the model’s release.

Qwen
Own leaderboard Not tested since Sep 2025
Qwen2.5-VL-7B 55.6%

Source: Official ODV-Bench leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Seed2.0 Pro Qwen2.5-VL-7B LLaVA-OneVision 7B 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Seed2.0 Pro Qwen2.5-VL-7B

Sources: Official ODV-Bench leaderboard and labs’ model cards

Labs headline it less often now, last in April 2026

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

ByteDance Seed
DateLaunch postScoreAs the post writes it
Apr 28, 2026 Seed2.0 Lite (0428) 66 ODVBench

One system card reports it

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
Feb 14, 2026 Seed2.0 Pro model card 72.5 ODVBench (Table 9) Capabilities

More in Audio and video

All 233 in Audio and video
  • Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy. Qwen 3.8 Max 88.7%
  • LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization. Gemini 4 Argon 91.7%
  • MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy. Qwen 3.8 Max 82.4%
  • MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy. Qwen 3.8 Max 90.8%
  • FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better. Phi-4-multimodal 4.00%
  • VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules. Qwen3.5-Omni-Plus 93.1%