K9-Bench
Models answer 4,744 questions about dog behavior in 907 videos. The evaluator compares each free-form answer with five options that the model never sees.
The authors rank 11 model variants on the published accuracy table. GPT-4o grades the same responses separately for qualities such as factual accuracy.
- Made by
- Ogmen Robotics and Georgia Institute of Technology
- Released
- June 2026
- Kind
- Video understanding
- Level
- School
- Scored
- Checked automatically
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Subject
- Biological sciences
- Language
- English
- Interaction
- Single turn
- Items from
- Real work and use, Generated
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision, Biology and medicine
- Strongest model tested
-
Gemini 2.5 Pro Jun 2025 #97 on the capability index
- Last model added
- October 2025
- Scores refreshed
- November 20th, 2026
Gemini 2.5 Pro leads with 40.1% among models released by October 2025 Gemini 2.5 Pro leads with 40.1% among models released by October 2025
Best score per entry from the benchmark's own leaderboard, read on November 20th, 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. This view shows only models we track, each at its best.
The newest model with a score here came out in October 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1
Gemini 2.5 Pro Jun 2025, 1 FPS 40.1%
- 2
Qwen3-VL-235B-A22B 1 FPS 36.5%
- 3 InternVL-14B 32 frames 33%
- 4 InternVL-8B 32 frames 32.2%
- 5 Qwen3-VL-32B 32 frames 31.4%
- 6 GPT-4o Mar 2025, 32 frames 30.8%
- 7
Qwen3-VL-32B-Thinking 32 frames 30.7%
- 8 Qwen3-VL-8B 32 frames 28.8%
- 9 Qwen3-VL-4B 32 frames 27.6%
- 10
Qwen3-VL-8B-Thinking 32 frames 25.3%
- 11
Qwen3-VL-4B-Thinking 32 frames 22.4%
Source: Official K9-Bench leaderboard
- 1
Gemini 2.5 Pro Jun 2025, 1 FPS 40.1%
- 2
Qwen3-VL-235B-A22B 1 FPS 36.5%
- 3 GPT-4o Mar 2025, 32 frames 30.8%
- 4
Qwen3-VL-32B-Thinking 32 frames 30.7%
- 5
Qwen3-VL-8B-Thinking 32 frames 25.3%
- 6
Qwen3-VL-4B-Thinking 32 frames 22.4%
Source: Official K9-Bench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 40.1% | 1 FPS | Jun 2025 | Official leaderboard |
| 2 | | 36.5% | 1 FPS | Sep 2025 | Official leaderboard |
| 3 | InternVL-14B | 33% | 32 frames | Official leaderboard | |
| 4 | InternVL-8B | 32.2% | 32 frames | Official leaderboard | |
| 5 | Qwen3-VL-32B | 31.4% | 32 frames | Official leaderboard | |
| 6 | GPT-4o Mar 2025 | 30.8% | 32 frames | Mar 2025 | Official leaderboard |
| 7 | | 30.7% | 32 frames | Oct 2025 | Official leaderboard |
| 8 | Qwen3-VL-8B | 28.8% | 32 frames | Official leaderboard | |
| 9 | Qwen3-VL-4B | 27.6% | 32 frames | Official leaderboard | |
| 10 | | 25.3% | 32 frames | Oct 2025 | Official leaderboard |
| 11 | | 22.4% | 32 frames | Oct 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 40.1% | 1 FPS | Jun 2025 | Official leaderboard |
| 2 | | 36.5% | 1 FPS | Sep 2025 | Official leaderboard |
| 3 | GPT-4o Mar 2025 | 30.8% | 32 frames | Mar 2025 | Official leaderboard |
| 4 | | 30.7% | 32 frames | Oct 2025 | Official leaderboard |
| 5 | | 25.3% | 32 frames | Oct 2025 | Official leaderboard |
| 6 | | 22.4% | 32 frames | Oct 2025 | Official leaderboard |
Rankings by part
We read 5 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Posture Analysis, Action Sequence and Context Analysis. Gemini 2.5 Pro leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.
Posture Analysis
6 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 37.7%
- 2
Qwen3-VL-235B-A22B 34.2%
- 3 GPT-4o Mar 2025 29.6%
- 4
Qwen3-VL-32B-Thinking 27.8%
- 5
Qwen3-VL-8B-Thinking 24%
- 6
Qwen3-VL-4B-Thinking 21.2%
Action Sequence
6 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 36%
- 2
Qwen3-VL-235B-A22B 33.5%
- 3
Qwen3-VL-32B-Thinking ↑1 28.8%
- 4 GPT-4o Mar 2025 ↓1 26.3%
- 5
Qwen3-VL-8B-Thinking 21.8%
- 6
Qwen3-VL-4B-Thinking 18%
Context Analysis
6 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 44%
- 2
Qwen3-VL-235B-A22B 38.4%
- 3 GPT-4o Mar 2025 35%
- 4
Qwen3-VL-32B-Thinking 33.7%
- 5
Qwen3-VL-8B-Thinking 29.7%
- 6
Qwen3-VL-4B-Thinking 28%
Cause-Effect Analysis
6 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 44.7%
- 2
Qwen3-VL-235B-A22B 39.9%
- 3 GPT-4o Mar 2025 33.9%
- 4
Qwen3-VL-32B-Thinking 32.8%
- 5
Qwen3-VL-8B-Thinking 24.9%
- 6
Qwen3-VL-4B-Thinking 23.1%
Interaction Analysis
6 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 42.3%
- 2
Qwen3-VL-235B-A22B 39.4%
- 3 GPT-4o Mar 2025 33.9%
- 4
Qwen3-VL-32B-Thinking 32.1%
- 5
Qwen3-VL-8B-Thinking 29.3%
- 6
Qwen3-VL-4B-Thinking 26.2%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official K9-Bench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official K9-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
- 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
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 13 months ago, in October 2025, for Qwen3-VL-32B-Thinking. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (11 entries).
- Pass Models scored 6 models we track, plus 5 other entries on its leaderboard.
- Warning Newest score The newest came 13 months ago, in October 2025, for Qwen3-VL-32B-Thinking.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 2.5 Pro (#97).
- Pass Headroom The best is 40.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on ogmenrobotics.github.io: 6 rows for models we track and 5 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.