Researchers and developers needing a small image-understanding model they can fine-tune; free under Apache 2.0, benchmarks are StepFun's own.
Kenning's take
What it does
Step3-VL-10B is a 10-billion-parameter open-source vision-language model that answers questions about images, reads text in pictures and documents, and handles maths and app screenshots. StepFun says it matches or beats open models 10 to 20 times its size on its benchmarks. Base and instruction-tuned weights are free on Hugging Face under Apache 2.0.
- 10 billion parameters
- Open base and instruction-tuned weights
- STEM reasoning, OCR, GUI and spatial understanding
- PaCoRe parallel reasoning mode
- Full fine-tuning and custom deployment
Who it suits. Developers and researchers who need a small image-understanding model they can run and fine-tune themselves.
Key facts
- Category
- AI models
- Made by
- StepFun
- Based in
- China
- Released
- 13 January 2026
- How to get it
- free open weights on Hugging Face and ModelScope
- Works on
- Hugging Face (open weights), ModelScope
- In the UK
- Yes
- Free plan
- Free open weights under Apache 2.0
FAQ
When was Step3-VL-10B released?
Its weights were published on Hugging Face on 13 January 2026.
Is Step3-VL-10B free?
Yes, the weights are free to download under the Apache 2.0 licence.
What can Step3-VL-10B do?
Answer questions about images, read text and documents, handle maths problems and understand app screenshots.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.