MiniCPM-V 4.5
MiniCPM-V 4.5 scores 17.1 out of 100 overall, which puts it at rank 33 of 36 models on the AI Vision Index. It is strongest at Not being fooled (40) and weakest at Space, 3D and motion (6). It beats the average of all 36 models in none of the six categories.
Overall score17.195% interval 15.9–18.7
Rank33of 36 models
Cost per 1,000 answers$0 (local)
Seconds per answer1.3 s on a laptop GPU
Thinking effortdefault
Maker OpenBMB
WeightsOpen weights
AccessRuns locally
Tested
Reading21
Counting12
Space, 3D and motion6
Spotting changes and errors16
Diagrams7
Not being fooled40
The bar is the score; the light band is the 95% interval. The overall score is the average of the six.
Reading
Reading a clock2
Reading a meter16
Eye chart0
Reading a label55
Reading a chart31
Test strips27
Adding up coins0
Traffic light from a car38
Counting
Counting objects10
Counting stacked goods14
Counting on a conveyor13
Space, 3D and motion
Which is nearer30
Which way it faces9
Same or mirror image0
Will it topple0
Match the render0
Which bay can I park in?0
Follow the hidden object5
Spotting changes and errors
Spot the change38
Shelf against plan8
Rendering glitch12
Check the invoice6
Diagrams
Follow the cable18
Follow the pipe2
Missing dimension1
Not being fooled
Absent objects82
Saying "I can't tell"62
Optical illusions3
Real or AI-made12
Scores are corrected for guessing: 0 is no better than a guess, 100 is always right. All models on every test
Neighbours on the leaderboard
- Qwen3-VL-8B 26.0 Compare
- Gemma 4 E4B 13.7 Compare
Badge for your site
HTML
<a href="https://aivisionindex.com/models/minicpm-v-4-5/"><img src="https://aivisionindex.com/badge/minicpm-v-4-5.svg" alt="MiniCPM-V 4.5: ranked #33 of 36 on the AI Vision Index"></a>Markdown
[](https://aivisionindex.com/models/minicpm-v-4-5/)Results are CC BY 4.0, so you are free to reuse them if you credit the AI Vision Index. All results (JSON)