PCA-Bench
Models interpret visual scenes, explain their reasoning, and select actions for embodied tasks. GPT-4-0125 compares perception concepts, reasoning, and actions with annotated references.
The leaderboard ranks action scores and Genuine PCA scores. Local multiple-choice accuracy is a separate metric.
- Made by
- Peking University and Alibaba Group
- Released
- February 2024
- Kind
- Embodied reasoning
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision
- Last model added
- March 2024
- Scores refreshed
- November 21st, 2026
GPT4-Vision-1106 tops the leaderboard with 0.79 Qwen-VL-Max leads with 0.70 among models released by March 2024
Best score per entry from the benchmark's own leaderboard, read on November 21st, 2026. 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. The best model we track is Qwen-VL-Max, at 0.70. This view shows only models we track, each at its best.
The newest model with a score here came out in March 2024, 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 GPT4-Vision-1106 open 0.79
- 2 GPT4-Vision-1106-Preview closed 0.72
- 3 Qwen-VL-Max closed 0.70
- 4 Gemini Pro Vision open 0.64
- 5
LLaVA-v1.5-13B closed 0.57
- 6
Yi-VL-34B closed 0.56
- 7
DeepSeek-VL-7B-Chat open 0.51
- 8 Qwen-VL-Chat closed 0.49
- 9 LLaVA-1.5 7B closed 0.45
- 10
Yi-VL-6B closed 0.44
Source: Official PCA-Bench leaderboard
- 1 Qwen-VL-Max closed 0.70
- 2
LLaVA-v1.5-13B closed 0.57
- 3
Yi-VL-34B closed 0.56
- 4
DeepSeek-VL-7B-Chat open 0.51
- 5 Qwen-VL-Chat closed 0.49
- 6
Yi-VL-6B closed 0.44
Source: Official PCA-Bench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT4-Vision-1106 | 0.79 | open | Official leaderboard | |
| 2 | GPT4-Vision-1106-Preview | 0.72 | closed | Official leaderboard | |
| 3 | Qwen-VL-Max | 0.70 | closed | Jan 2024 | Official leaderboard |
| 4 | Gemini Pro Vision | 0.64 | open | Official leaderboard | |
| 5 | | 0.57 | closed | Oct 2023 | Official leaderboard |
| 6 | | 0.56 | closed | Dec 2023 | Official leaderboard |
| 7 | | 0.51 | open | Mar 2024 | Official leaderboard |
| 8 | Qwen-VL-Chat | 0.49 | closed | Aug 2023 | Official leaderboard |
| 9 | LLaVA-1.5 7B | 0.45 | closed | Official leaderboard | |
| 10 | | 0.44 | closed | Dec 2023 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Qwen-VL-Max | 0.70 | closed | Jan 2024 | Official leaderboard |
| 2 | | 0.57 | closed | Oct 2023 | Official leaderboard |
| 3 | | 0.56 | closed | Dec 2023 | Official leaderboard |
| 4 | | 0.51 | open | Mar 2024 | Official leaderboard |
| 5 | Qwen-VL-Chat | 0.49 | closed | Aug 2023 | Official leaderboard |
| 6 | | 0.44 | closed | Dec 2023 | Official leaderboard |
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.
Source: Official PCA-Bench leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Qwen-VL-Max | closed | 0.70 | |||
| Qwen-VL-Max | open | 0.64 | |||
| closed | 0.57 | ||||
| closed | 0.56 | ||||
| open | 0.55 | ||||
| open | 0.51 | ||||
| open | 0.50 | ||||
| Qwen-VL-Chat | closed | 0.49 | |||
| closed | 0.49 | ||||
| closed | 0.44 | ||||
| open | 0.43 | ||||
| Qwen-VL-Chat | open | 0.40 |
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.
Source: Official PCA-Bench leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Qwen-VL-Max | closed | 0.70 | |||
| Qwen-VL-Max | open | 0.64 | |||
| closed | 0.57 | ||||
| closed | 0.56 | ||||
| open | 0.55 | ||||
| open | 0.51 | ||||
| open | 0.50 | ||||
| Qwen-VL-Chat | closed | 0.49 | |||
| closed | 0.49 | ||||
| closed | 0.44 | ||||
| open | 0.43 | ||||
| Qwen-VL-Chat | open | 0.40 |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official PCA-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 PCA-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 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 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 2 years ago, in March 2024, for DeepSeek-VL-7B-Chat. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (10 entries).
- Pass Models scored 6 models we track, plus 4 other entries on its leaderboard.
- Warning Newest score The newest came 2 years ago, in March 2024, for DeepSeek-VL-7B-Chat.
- Note Frontier models None of its models are on the capabilities index.
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 12 rows for models we track and 4 other entries at the last read.
- Pass Last read November 21st, 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.