Benchmark Almanac

RVG-Bench

Qwen3.8-Omni-Flash 49.6% leads since October 2026
Not in model cards

Models locate the video interval that answers a question about sounds and images. The evaluator measures overlap with the annotated interval.

Made by
Yue Feng and colleagues
Released
October 2026
Kind
Audio understanding and voice
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Video, Audio → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
September 2026
Scores refreshed
November 20th, 2026
22
Entries on its own leaderboard, 9 models we track

Qwen3.8-Omni-Flash leads with 49.6% among models released by September 2026 Qwen3.8-Omni-Flash leads with 49.6% among models released by September 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The score is overall mean intersection over union, expressed as a percentage. 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. This view shows only models we track, each at its best.

The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Qwen3.8-Omni-Flash Chain of thought, audio-visual input 49.6%
  2. 2 Seed2.0 Lite (0215) Chain of thought, audio-visual input 47%
  3. 3 Gemini 3 Pro Chain of thought, audio-visual input 44.4%
  4. 4 Seed2.0 Pro Chain of thought, visual input 41.9%
  5. 5 GPT-5.5 Chain of thought, visual input 38.5%
  6. 6 Qwen3.8 Flash Chain of thought, visual input 38.2%
  7. 7 TimeLens-8B Chain of thought, visual input 30.1%
  8. 8 Qwen3-VL-8B Chain of thought, visual input 26.3%
  9. 9 ChronusOmni Chain of thought, audio-visual input 23.4%
  10. 10 TimeLens-7B Chain of thought, visual input 21.8%
  11. 11 Time-R1-7B Chain of thought, visual input 19.7%
  12. 12 TRACE Chain of thought, visual input 14.1%
  13. 13 Qwen2.5-VL-7B Chain of thought, visual input 13.4%
  14. 14 VideoChat-R1 Chain of thought, visual input 11.9%
  15. 15 Qwen3-Omni Chain of thought, audio-visual input 10.7%
  16. 16 TimeSuite Chain of thought, visual input 9.40%
  17. 17 Video-R1-7B Chain of thought, visual input 6.40%
  18. 18 HumanOmniV2 Chain of thought, audio-visual input 5.10%
  19. 19 Qwen2.5-Omni Chain of thought, audio-visual input 5.10%
  20. 20 Ola-7B Chain of thought, audio-visual input 4.50%
  21. 21 MiniCPM-o-2_6 Chain of thought, audio-visual input 2.20%
  22. 22 MiMo-VL-7B-RL Chain of thought, visual input 1.70%

Source: Official RVG-Bench leaderboard

#EntryScoreSettingDateSource
1 Qwen3.8-Omni-Flash 49.6% Chain of thought, audio-visual input Sep 2026 Official leaderboard
2 Seed2.0 Lite (0215) 47% Chain of thought, audio-visual input Feb 2026 Official leaderboard
3 Gemini 3 Pro 44.4% Chain of thought, audio-visual input Nov 2025 Official leaderboard
4 Seed2.0 Pro 41.9% Chain of thought, visual input Feb 2026 Official leaderboard
5 GPT-5.5 38.5% Chain of thought, visual input Apr 2026 Official leaderboard
6 Qwen3.8 Flash 38.2% Chain of thought, visual input Aug 2026 Official leaderboard
7 TimeLens-8B 30.1% Chain of thought, visual input Official leaderboard
8 Qwen3-VL-8B 26.3% Chain of thought, visual input Official leaderboard
9 ChronusOmni 23.4% Chain of thought, audio-visual input Official leaderboard
10 TimeLens-7B 21.8% Chain of thought, visual input Official leaderboard
11 Time-R1-7B 19.7% Chain of thought, visual input Official leaderboard
12 TRACE 14.1% Chain of thought, visual input Official leaderboard
13 Qwen2.5-VL-7B 13.4% Chain of thought, visual input Jan 2025 Official leaderboard
14 VideoChat-R1 11.9% Chain of thought, visual input Official leaderboard
15 Qwen3-Omni 10.7% Chain of thought, audio-visual input Official leaderboard
16 TimeSuite 9.40% Chain of thought, visual input Official leaderboard
17 Video-R1-7B 6.40% Chain of thought, visual input Official leaderboard
18 HumanOmniV2 5.10% Chain of thought, audio-visual input Official leaderboard
19 Qwen2.5-Omni 5.10% Chain of thought, audio-visual input Official leaderboard
20 Ola-7B 4.50% Chain of thought, audio-visual input Official leaderboard
21 MiniCPM-o-2_6 2.20% Chain of thought, audio-visual input Jan 2025 Official leaderboard
22 MiMo-VL-7B-RL 1.70% Chain of thought, visual input May 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Qwen3.8-Omni-Flash 49.6% Chain of thought, audio-visual input Sep 2026 Official leaderboard
2 Seed2.0 Lite (0215) 47% Chain of thought, audio-visual input Feb 2026 Official leaderboard
3 Gemini 3 Pro 44.4% Chain of thought, audio-visual input Nov 2025 Official leaderboard
4 Seed2.0 Pro 41.9% Chain of thought, visual input Feb 2026 Official leaderboard
5 GPT-5.5 38.5% Chain of thought, visual input Apr 2026 Official leaderboard
6 Qwen3.8 Flash 38.2% Chain of thought, visual input Aug 2026 Official leaderboard
7 Qwen2.5-VL-7B 13.4% Chain of thought, visual input Jan 2025 Official leaderboard
8 MiniCPM-o-2_6 2.20% Chain of thought, audio-visual input Jan 2025 Official leaderboard
9 MiMo-VL-7B-RL 1.70% Chain of thought, visual input May 2025 Official leaderboard

Rankings by part

We read 10 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Causal, Sequential and Referential. Qwen3.8-Omni-Flash leads 7 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Causal

9 models we track have a score here.

  1. 1 Qwen3.8-Omni-Flash 38.5%
  2. 2 Seed2.0 Lite (0215) 36.8%
  3. 3 Gemini 3 Pro 31.3%
  4. 4 Seed2.0 Pro 28.9%
  5. 5 GPT-5.5 28.2%
  6. 6 Qwen3.8 Flash 26.5%
  7. 7 Qwen2.5-VL-7B 11.2%
  8. 8 MiMo-VL-7B-RL ↑1 2.60%
  9. 9 MiniCPM-o-2_6 ↓1 1.30%

Number one over time

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

Qwen
Own leaderboard Not tested since Oct 2026

Source: Official RVG-Bench leaderboard

Scores over time

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

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Qwen3.8-Omni-Flash Seed2.0 Lite (0215) Qwen2.5-VL-7B MiniCPM-o-2_6 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Qwen3.8-Omni-Flash Qwen2.5-VL-7B

Source: Official RVG-Bench leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Audio and video

All 695 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%

Data audit

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (22 entries).
  • Pass Models scored 9 models we track, plus 13 other entries on its leaderboard.
  • Pass Newest score The newest is from September 2026, for Qwen3.8-Omni-Flash.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
  • Pass Headroom The best is 49.6%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on fyrisk.github.io: 9 rows for models we track and 13 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check