Benchmark Almanac

DashBench

GPT-5 34.8% leads since November 2026
Not in model cards

Models explain rare events in driving videos, including near misses and collisions. The authors compare 25 video models on 1,709 clips.

Released
November 2026
Kind
Video understanding
Level
Everyday
Scored
Graded by a model
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Video → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
In views
Vision
Strongest model tested
Gemini 3 Flash #60 on the capability index
Last model added
December 2025
Scores refreshed
November 20th, 2026
25
Entries on its own leaderboard, 11 models we track

GPT-5 leads with 34.8% among models released by December 2025 GPT-5 leads with 34.8% among models released by December 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 December 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 GPT-5 34.8%
  2. 2 Gemini 3 Flash 34.4%
  3. 3 Qwen3-VL-235B-A22B 20.7%
  4. 4 Qwen3-VL (8B) 18.5%
  5. 5 Qwen3-VL (32B) 17.2%
  6. 6 Ovis2.5-9B 16.4%
  7. 7 InternVL3.5-241B-A28B 16.1%
  8. 8 InternVL3.5-38B 12.8%
  9. 9 Tarsier2 (7B) 12.6%
  10. 10 MiniCPM-V 4.5 12.3%
  11. 11 Keye-VL-8B-Preview 11.4%
  12. 12 LLaVA-OV-RS (7B) 10%
  13. 13 InternVL3.5-8B 8.64%
  14. 14 InternVideo2.5 (8B) 7.76%
  15. 15 IXC-2.5 (7B) 7.74%
  16. 16 GLM-4.1V-9B-Thinking 6.33%
  17. 17 CrashChat (7B) 5.91%
  18. 18 VITA-1.5 5.44%
  19. 19 LLaVA-OV-1.5 (8B) 4.38%
  20. 20 DriveLMM-o1 (8B) 2.76%
  21. 21 LLaVA-Next-Video (7B) 1.57%
  22. 22 DriveMM (8B) 1.42%
  23. 23 MiniInternVL2-BDD (4B) 1.10%
  24. 24 MiniInternVL2-DLM (4B) 0.87%
  25. 25 Dolphins (7B) 0.32%

Source: Official DashBench leaderboard

#EntryScoreSettingDateSource
1 GPT-5 34.8% Aug 2025 Official leaderboard
2 Gemini 3 Flash 34.4% Dec 2025 Official leaderboard
3 Qwen3-VL-235B-A22B 20.7% Sep 2025 Official leaderboard
4 Qwen3-VL (8B) open weights 18.5% Official leaderboard
5 Qwen3-VL (32B) open weights 17.2% Official leaderboard
6 Ovis2.5-9B 16.4% Aug 2025 Official leaderboard
7 InternVL3.5-241B-A28B 16.1% Aug 2025 Official leaderboard
8 InternVL3.5-38B 12.8% Aug 2025 Official leaderboard
9 Tarsier2 (7B) open weights 12.6% Official leaderboard
10 MiniCPM-V 4.5 12.3% Aug 2025 Official leaderboard
11 Keye-VL-8B-Preview 11.4% Jun 2025 Official leaderboard
12 LLaVA-OV-RS (7B) 10% Official leaderboard
13 InternVL3.5-8B 8.64% Aug 2025 Official leaderboard
14 InternVideo2.5 (8B) open weights 7.76% Official leaderboard
15 IXC-2.5 (7B) open weights 7.74% Official leaderboard
16 GLM-4.1V-9B-Thinking 6.33% Jul 2025 Official leaderboard
17 CrashChat (7B) 5.91% Official leaderboard
18 VITA-1.5 5.44% Dec 2024 Official leaderboard
19 LLaVA-OV-1.5 (8B) open weights 4.38% Official leaderboard
20 DriveLMM-o1 (8B) 2.76% Official leaderboard
21 LLaVA-Next-Video (7B) open weights 1.57% Official leaderboard
22 DriveMM (8B) 1.42% Official leaderboard
23 MiniInternVL2-BDD (4B) 1.10% Official leaderboard
24 MiniInternVL2-DLM (4B) 0.87% Official leaderboard
25 Dolphins (7B) 0.32% Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5 34.8% Aug 2025 Official leaderboard
2 Gemini 3 Flash 34.4% Dec 2025 Official leaderboard
3 Qwen3-VL-235B-A22B 20.7% Sep 2025 Official leaderboard
4 Ovis2.5-9B 16.4% Aug 2025 Official leaderboard
5 InternVL3.5-241B-A28B 16.1% Aug 2025 Official leaderboard
6 InternVL3.5-38B 12.8% Aug 2025 Official leaderboard
7 MiniCPM-V 4.5 12.3% Aug 2025 Official leaderboard
8 Keye-VL-8B-Preview 11.4% Jun 2025 Official leaderboard
9 InternVL3.5-8B 8.64% Aug 2025 Official leaderboard
10 GLM-4.1V-9B-Thinking 6.33% Jul 2025 Official leaderboard
11 VITA-1.5 5.44% Dec 2024 Official leaderboard

Rankings by part

We read 6 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Q1 (Unusual), Q2 (Dangerous) and Q3 (Describe). GPT-5 leads 5 of them, and Gemini 3 Flash leads the other one. Arrows show how far a model moved from its overall rank among the models we track.

Q1 (Unusual)

11 models we track have a score here.

  1. 1 Gemini 3 Flash ↑1 32.4%
  2. 2 GPT-5 ↓1 19.9%
  3. 3 Ovis2.5-9B ↑1 19.7%
  4. 4 InternVL3.5-241B-A28B ↑1 19.3%
  5. 5 Qwen3-VL-235B-A22B ↓2 18.7%
  6. 6 MiniCPM-V 4.5 ↑1 13.3%
  7. 7 InternVL3.5-38B ↓1 12.6%
  8. 8 Keye-VL-8B-Preview 12.4%
  9. 9 GLM-4.1V-9B-Thinking ↑1 7.85%
  10. 10 VITA-1.5 ↑1 7.31%

Number one over time

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

OpenAI
Own leaderboard 1 leader
GPT-5 34.8%

Source: Official DashBench 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%) GPT-5 Keye-VL-8B-Preview VITA-1.5 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-5 Keye-VL-8B-Preview

Source: Official DashBench 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

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 11 months ago, in December 2025, for Gemini 3 Flash. 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 (25 entries).
  • Pass Models scored 11 models we track, plus 14 other entries on its leaderboard.
  • Warning Newest score The newest came 11 months ago, in December 2025, for Gemini 3 Flash.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Flash (#60).
  • Pass Headroom The best is 34.8%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on cornercase-dashbench.github.io: 11 rows for models we track and 14 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