Benchmark Almanac

MVDriveBench

GPT-5.5 80.1% leads since November 2026
Not in model cards

Models answer driving questions using synchronized videos from cameras facing different directions. The authors compare surround-view results with front-camera results.

Made by
Shunming Liu and colleagues
Released
November 2026
Kind
Embodied reasoning
Level
Everyday
Scored
Checked automatically
Answers
Multiple choice
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
A host
In views
Vision
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
April 2026
Scores refreshed
November 20th, 2026
23
Entries in its paper, 15 models we track

GPT-5.5 leads with 80.1% among models released by April 2026 GPT-5.5 leads with 80.1% among models released by April 2026

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. 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 April 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 GPT-5.5 Surround-view 80.1%
  2. 2 GPT-5 Surround-view 78.7%
  3. 3 Gemini 3 Pro Front-view 78.4%
  4. 4 Gemini-2.5-Pro (snapshot unspecified) Surround-view 76.7%
  5. 5 Claude Opus 4.6 Surround-view 75.9%
  6. 6 Qwen3.5 397B-A17B Surround-view 74.7%
  7. 7 GPT-4o (snapshot unspecified) Surround-view 74.3%
  8. 8 Qwen2.5-VL-32B Surround-view 71.5%
  9. 9 InternVL3-78B Surround-view 71.2%
  10. 10 InternVL3-38B Surround-view 70.9%
  11. 11 Qwen2.5-VL-72B Surround-view 69.4%
  12. 12 Qwen3-VL-30B-A3B-Thinking Surround-view 66.4%
  13. 13 Video-R1-7B Surround-view 62.2%
  14. 14 Qwen3.5-27B Surround-view 61.7%
  15. 15 VGLM-QA-VGGT-8B Surround-view 56.4%
  16. 16 MiMo-VL-7B-SFT Surround-view 54.4%
  17. 17 Cosmos-Reason2-8B Surround-view 51.3%
  18. 18 LLaVA-OneVision 7B Surround-view 50.5%
  19. 19 ERNIE-4.5-VL-28B-A3B-Thinking Surround-view 49.9%
  20. 20 MiMoEmbodied-7B Surround-view 47.3%
  21. 21 ReCogDrive-VLM-8B Surround-view 47.3%
  22. 22 GLM-4.1V-9B-Thinking Front-view 47.2%
  23. 23 Cosmos-Reason1-7B Surround-view 45.7%

Source: The MVDriveBench paper

#EntryScoreSettingDateSource
1 GPT-5.5 80.1% Surround-view Apr 2026 Official leaderboard
2 GPT-5 78.7% Surround-view Aug 2025 Official leaderboard
3 Gemini 3 Pro 78.4% Front-view Nov 2025 Official leaderboard
4 Gemini-2.5-Pro (snapshot unspecified) 76.7% Surround-view Official leaderboard
5 Claude Opus 4.6 75.9% Surround-view Feb 2026 Official leaderboard
6 Qwen3.5 397B-A17B 74.7% Surround-view Feb 2026 Official leaderboard
7 GPT-4o (snapshot unspecified) 74.3% Surround-view Official leaderboard
8 Qwen2.5-VL-32B 71.5% Surround-view Mar 2025 Official leaderboard
9 InternVL3-78B 71.2% Surround-view Apr 2025 Official leaderboard
10 InternVL3-38B 70.9% Surround-view Apr 2025 Official leaderboard
11 Qwen2.5-VL-72B 69.4% Surround-view Jan 2025 Official leaderboard
12 Qwen3-VL-30B-A3B-Thinking 66.4% Surround-view Oct 2025 Official leaderboard
13 Video-R1-7B 62.2% Surround-view Official leaderboard
14 Qwen3.5-27B 61.7% Surround-view Feb 2026 Official leaderboard
15 VGLM-QA-VGGT-8B 56.4% Surround-view Official leaderboard
16 MiMo-VL-7B-SFT 54.4% Surround-view May 2025 Official leaderboard
17 Cosmos-Reason2-8B 51.3% Surround-view Official leaderboard
18 LLaVA-OneVision 7B 50.5% Surround-view Aug 2024 Official leaderboard
19 ERNIE-4.5-VL-28B-A3B-Thinking 49.9% Surround-view Nov 2025 Official leaderboard
20 MiMoEmbodied-7B 47.3% Surround-view Official leaderboard
21 ReCogDrive-VLM-8B 47.3% Surround-view Official leaderboard
22 GLM-4.1V-9B-Thinking 47.2% Front-view Jul 2025 Official leaderboard
23 Cosmos-Reason1-7B 45.7% Surround-view Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5.5 80.1% Surround-view Apr 2026 Official leaderboard
2 GPT-5 78.7% Surround-view Aug 2025 Official leaderboard
3 Gemini 3 Pro 78.4% Front-view Nov 2025 Official leaderboard
4 Claude Opus 4.6 75.9% Surround-view Feb 2026 Official leaderboard
5 Qwen3.5 397B-A17B 74.7% Surround-view Feb 2026 Official leaderboard
6 Qwen2.5-VL-32B 71.5% Surround-view Mar 2025 Official leaderboard
7 InternVL3-78B 71.2% Surround-view Apr 2025 Official leaderboard
8 InternVL3-38B 70.9% Surround-view Apr 2025 Official leaderboard
9 Qwen2.5-VL-72B 69.4% Surround-view Jan 2025 Official leaderboard
10 Qwen3-VL-30B-A3B-Thinking 66.4% Surround-view Oct 2025 Official leaderboard
11 Qwen3.5-27B 61.7% Surround-view Feb 2026 Official leaderboard
12 MiMo-VL-7B-SFT 54.4% Surround-view May 2025 Official leaderboard
13 LLaVA-OneVision 7B 50.5% Surround-view Aug 2024 Official leaderboard
14 ERNIE-4.5-VL-28B-A3B-Thinking 49.9% Surround-view Nov 2025 Official leaderboard
15 GLM-4.1V-9B-Thinking 47.2% Front-view Jul 2025 Official leaderboard

Rankings by part

We read 15 parts from the benchmark’s paper, each ranking models on one part of the test, such as Rel. Mot., Obj. Mot. and Tracking. GPT-5.5 leads 5 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Rel. Mot.

15 models we track have a score here.

  1. 1 GPT-5 ↑1 64.9%
  2. 2 Qwen3.5-27B ↑9 63.2%
  3. 3 GPT-5.5 ↓2 62.3%
  4. 4 Gemini 3 Pro ↓1 61.3%
  5. 5 Claude Opus 4.6 ↓1 60.9%
  6. 6 Qwen3.5 397B-A17B ↓1 57.9%
  7. 7 InternVL3-38B ↑1 53.2%
  8. 8 Qwen2.5-VL-32B ↓2 49.9%
  9. 9 Qwen3-VL-30B-A3B-Thinking ↑1 49%
  10. 10 InternVL3-78B ↓3 48.7%

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
GPT-5.5
Front-view
80.1%
GPT-5
Front-view
78.7%
Gemini 3 Pro
Surround-vi…
78.4%
Claude Opus 4.6
Front-view
75.9%
Qwen3.5 397B-A17B
Front-view Surround-vi…
74.7%
Qwen2.5-VL-32B
Front-view Surround-vi…
71.5%
InternVL3-78B
Front-view Surround-vi…
71.2%
InternVL3-38B
Front-view Surround-vi…
70.9%
Qwen2.5-VL-72B
Front-view Surround-vi…
69.4%
Qwen3-VL-30B-A3B-Thinking
Front-view Surround-vi…
66.4%
Qwen3.5-27B
Front-view Surround-vi…
61.7%
MiMo-VL-7B-SFT
Front-view Surround-vi…
54.4%

Source: The MVDriveBench paper

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5.5 Surround-view 80.1%
GPT-5.5 Front-view 78.8%
GPT-5 Surround-view 78.7%
Gemini 3 Pro Front-view 78.4%
Gemini 3 Pro Surround-view 78.3%
GPT-5 Front-view 77.2%
Claude Opus 4.6 Surround-view 75.9%
Qwen3.5 397B-A17B Surround-view 74.7%
Claude Opus 4.6 Front-view 73.9%
Qwen2.5-VL-32B Surround-view 71.5%
InternVL3-78B Surround-view 71.2%
InternVL3-38B Surround-view 70.9%
Qwen2.5-VL-72B Surround-view 69.4%
Qwen3-VL-30B-A3B-Thinking Surround-view 66.4%
Qwen3.5-27B Surround-view 61.7%
Qwen3.5 397B-A17B Front-view 59.6%
Qwen2.5-VL-32B Front-view 56.2%
InternVL3-38B Front-view 55.9%
InternVL3-78B Front-view 55.6%
MiMo-VL-7B-SFT Surround-view 54.4%
Qwen3-VL-30B-A3B-Thinking Front-view 53.6%
Qwen2.5-VL-72B Front-view 52.4%
LLaVA-OneVision 7B Surround-view 50.5%
ERNIE-4.5-VL-28B-A3B-Thinking Surround-view 49.9%
Qwen3.5-27B Front-view 47.5%
GLM-4.1V-9B-Thinking Front-view 47.2%
LLaVA-OneVision 7B Front-view 47.1%
MiMo-VL-7B-SFT Front-view 44.9%
ERNIE-4.5-VL-28B-A3B-Thinking Front-view 43.6%
GLM-4.1V-9B-Thinking Surround-view 43.2%

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
GPT-5.5
Front-view
80.1%
GPT-5
Front-view
78.7%
Gemini 3 Pro
Surround-vi…
78.4%
Claude Opus 4.6
Front-view
75.9%
Qwen3.5 397B-A17B
Front-view Surround-vi…
74.7%
Qwen2.5-VL-32B
Front-view Surround-vi…
71.5%
InternVL3-78B
Front-view Surround-vi…
71.2%
InternVL3-38B
Front-view Surround-vi…
70.9%
Qwen2.5-VL-72B
Front-view Surround-vi…
69.4%
Qwen3-VL-30B-A3B-Thinking
Front-view Surround-vi…
66.4%
Qwen3.5-27B
Front-view Surround-vi…
61.7%
MiMo-VL-7B-SFT
Front-view Surround-vi…
54.4%

Source: The MVDriveBench paper

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5.5 Surround-view 80.1%
GPT-5.5 Front-view 78.8%
GPT-5 Surround-view 78.7%
Gemini 3 Pro Front-view 78.4%
Gemini 3 Pro Surround-view 78.3%
GPT-5 Front-view 77.2%
Claude Opus 4.6 Surround-view 75.9%
Qwen3.5 397B-A17B Surround-view 74.7%
Claude Opus 4.6 Front-view 73.9%
Qwen2.5-VL-32B Surround-view 71.5%
InternVL3-78B Surround-view 71.2%
InternVL3-38B Surround-view 70.9%
Qwen2.5-VL-72B Surround-view 69.4%
Qwen3-VL-30B-A3B-Thinking Surround-view 66.4%
Qwen3.5-27B Surround-view 61.7%
Qwen3.5 397B-A17B Front-view 59.6%
Qwen2.5-VL-32B Front-view 56.2%
InternVL3-38B Front-view 55.9%
InternVL3-78B Front-view 55.6%
MiMo-VL-7B-SFT Surround-view 54.4%
Qwen3-VL-30B-A3B-Thinking Front-view 53.6%
Qwen2.5-VL-72B Front-view 52.4%
LLaVA-OneVision 7B Surround-view 50.5%
ERNIE-4.5-VL-28B-A3B-Thinking Surround-view 49.9%
Qwen3.5-27B Front-view 47.5%
GLM-4.1V-9B-Thinking Front-view 47.2%
LLaVA-OneVision 7B Front-view 47.1%
MiMo-VL-7B-SFT Front-view 44.9%
ERNIE-4.5-VL-28B-A3B-Thinking Front-view 43.6%
GLM-4.1V-9B-Thinking Surround-view 43.2%

Number one over time

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

OpenAI
Benchmark paper 1 leader
GPT-5.5 80.1%

Source: The MVDriveBench 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% 2025 2026 Near the ceiling (85%) GPT-5.5 GPT-5 Qwen2.5-VL-32B LLaVA-OneVision 7B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-5.5 Qwen2.5-VL-32B

Source: Official MVDriveBench 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 Robotics

All 172 in Robotics
  • ERQA In use 400 multiple-choice visual questions from robot and egocentric footage on spatial, trajectory, action and state reasoning. Accuracy. Gemini Robotics ER 2 78.5%
  • SimplerEnv In use Real-to-sim replicas of Google Robot and WidowX/Bridge setups. Policies scored by success rate on pick, move, drawer and Bridge tasks. Praxis-1 80%
  • RoboLab In use Zero-shot DROID Franka policies in Isaac Lab on 120 tabletop tasks at three instruction levels, 10 episodes each. Success rate. Praxis-1 46.2%
  • RoboDojo In use Generalist manipulation policies run 42 Isaac Sim tasks and 18 remote real-robot tasks. Scored by task score and success rate. GPT-6 Astra 27.6%
  • Progress classification (Google DeepMind) In use Models assign each frame of a robot video to one of five task-progress bands from 0 to 100 percent. Scored by accuracy. Gemini Robotics ER 2 57.4%
  • Human proximity (Google DeepMind robotics) In use Models judge from robot camera views whether a human is within 1 meter, a spatial awareness check for safe robot behavior. Scored by accuracy. Gemini Robotics ER 2 93%

Data audit

8 of 10 checks pass, 2 warnings. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark

Catalog record 1 warning 1 pass
  • Warning Task access We don’t know whether the tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (23 entries).
  • Pass Models scored 15 models we track, plus 8 other entries on its leaderboard.
  • Warning Newest score The newest came 7 months ago, in April 2026, for GPT-5.5.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
  • Pass Headroom The best is 80.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 30 rows for models we track and 8 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