Benchmark Almanac

VideoGAIA

Seed2.0 Pro 58.3% leads since August 2026
Not in model cards

Agents inspect selected video segments, search the web, and read pages before answering questions. Model judges check answers, and the harness reports accuracy by category and difficulty.

The released set contains 271 questions. The harness supports a fallback judge, which must be identified when reporting results.

Released
August 2026
Kind
Video understanding
Level
Everyday
Scored
Graded by a model
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Video → Text, Actions
Interaction
Agentic
Environment
Browser
Refresh
One fixed set
Run by
Anyone
In views
Agents, Vision
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
August 2026
Scores refreshed
November 18th, 2026

In the authors’ words

Video understanding is a fundamental task for evaluating the capabilities of multimodal large language models (MLLMs). However, existing leading models have already achieved approximately 90% accuracy on the Video-MME leaderboard, suggesting that conventional single-turn video understanding tasks are becoming increasingly saturated and insufficient for assessing the intelligence of advanced MLLMs. Towards this end, we introduce VideoGAIA, an agentic video understanding benchmark for general artificial intelligence (AI) assistants. Moving beyond one-shot video question answering, VideoGAIA formulates video understanding as a multi-turn, tool-augmented interaction process, where models must iteratively perceive videos, invoke external tools, gather complementary information, and integrate multimodal evidence across turns. VideoGAIA contains 271 model-human co-designed tasks covering diverse and complex real-world scenarios. Each video-question-answer instance is independently verified by three human experts to ensure both correctness and appropriate difficulty. All evaluated MLLMs, including frontier models such as GPT-5.5 and Kimi-K3, achieve less than 60% accuracy on VideoGAIA, highlighting its value as a high-quality and timely benchmark for evaluating next-generation MLLMs. We hope that VideoGAIA will facilitate the transition from conventional video understanding toward agentic video understanding.

Abstract of VideoGAIA: A Benchmark for General AI Assistants on Agentic Video Understanding (arXiv:2608.14718, CC0)
20
Models scored in its paper

Seed2.0 Pro leads with 58.3% among models released by August 2026

Best score per model from the benchmark's paper, read on November 18th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best.

The newest model with a score here came out in August 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 Seed2.0 Pro 58.3%
  2. 2 Qwen3.7-Plus 56.1%
  3. 3 Kimi K3 54.6%
  4. 4 GPT-5.5 53.1%
  5. 5 Kimi K2.6 52.8%
  6. 6 Qwen3.5-Plus 52%
  7. 7 GPT-5.4 52%
  8. 8 GPT-5.2 50.9%
  9. 9 Gemini 3.1 Pro 50.9%
  10. 10 Qwen3.5 397B-A17B 50.2%
  11. 11 GLM-5V-Turbo 50.2%
  12. 12 Qwen 3.8 Max 50.2%
  13. 13 Kimi K2.5 49.8%
  14. 14 Qwen 3.6 Plus 49.1%
  15. 15 MiMo-V2.5 48.7%
  16. 16 Claude Opus 4.6 48%
  17. 17 GLM-4.6V 46.9%
  18. 18 Gemini 3.5 Flash 43.9%
  19. 19 Claude Opus 4.8 36.9%
  20. 20 Claude Opus 4.7 34.7%

Source: The VideoGAIA paper

#ModelScoreSettingDateSource
1 Seed2.0 Pro 58.3% VideoGAIA agent loop Feb 2026 Official leaderboard
2 Qwen3.7-Plus 56.1% VideoGAIA agent loop Jun 2026 Official leaderboard
3 Kimi K3 54.6% VideoGAIA agent loop Jul 2026 Official leaderboard
4 GPT-5.5 53.1% VideoGAIA agent loop Apr 2026 Official leaderboard
5 Kimi K2.6 52.8% VideoGAIA agent loop Apr 2026 Official leaderboard
6 Qwen3.5-Plus 52% VideoGAIA agent loop Feb 2026 Official leaderboard
7 GPT-5.4 52% VideoGAIA agent loop Mar 2026 Official leaderboard
8 GPT-5.2 50.9% VideoGAIA agent loop Dec 2025 Official leaderboard
9 Gemini 3.1 Pro 50.9% VideoGAIA agent loop Feb 2026 Official leaderboard
10 Qwen3.5 397B-A17B 50.2% VideoGAIA agent loop Feb 2026 Official leaderboard
11 GLM-5V-Turbo 50.2% VideoGAIA agent loop Apr 2026 Official leaderboard
12 Qwen 3.8 Max 50.2% VideoGAIA agent loop Aug 2026 Official leaderboard
13 Kimi K2.5 49.8% VideoGAIA agent loop Jan 2026 Official leaderboard
14 Qwen 3.6 Plus 49.1% VideoGAIA agent loop Mar 2026 Official leaderboard
15 MiMo-V2.5 48.7% VideoGAIA agent loop Apr 2026 Official leaderboard
16 Claude Opus 4.6 48% VideoGAIA agent loop Feb 2026 Official leaderboard
17 GLM-4.6V 46.9% VideoGAIA agent loop Dec 2025 Official leaderboard
18 Gemini 3.5 Flash 43.9% VideoGAIA agent loop May 2026 Official leaderboard
19 Claude Opus 4.8 36.9% VideoGAIA agent loop May 2026 Official leaderboard
20 Claude Opus 4.7 34.7% VideoGAIA agent loop Apr 2026 Official leaderboard

Rankings by part

We read 6 parts from the benchmark’s paper, each ranking models on one part of the test, such as Culture, Geography and Daily life. GPT-5.4 leads 2 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Culture

20 models we track have a score here.

  1. 1 GPT-5.4 ↑6 56.9%
  2. 2 MiMo-V2.5 ↑13 53.5%
  3. 3 Seed2.0 Pro ↓2 51.7%
  4. 4 GLM-5V-Turbo ↑7 51.7%
  5. 5 GPT-5.2 ↑3 50%
  6. 6 Kimi K2.5 ↑7 50%
  7. 7 Kimi K2.6 ↓2 50%
  8. 8 GPT-5.5 ↓4 50%
  9. 9 Qwen3.7-Plus ↓7 50%
  10. 10 Kimi K3 ↓7 50%

Number one over time

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

ByteDance Seed
Benchmark paper Not tested since Aug 2026
Seed2.0 Pro 58.3%

Source: The VideoGAIA paper

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
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Seed2.0 Pro GPT-5.2 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Seed2.0 Pro GPT-5.2

Source: Official VideoGAIA 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 benchmark’s paper (20 entries).
  • Pass Models scored 20 models we track.
  • Pass Newest score The newest is from August 2026, for Qwen 3.8 Max.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
  • Pass Headroom The best is 58.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 20 rows for models we track at the last read.
  • Pass Last read November 18th, 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