step3
Step3 is our cutting-edge multimodal reasoning model—built on a Mixture-of-Experts architecture with 321B total parameters and 38B active.
text + image → text · made by stepfun-ai
Sold by
Nobody in the catalogue publishes a price for this yet.
About
step3 — a text model from stepfun-ai.
It takes text and images and returns text, with a context window of 65,536 tokens. It was published in July 2025, trained on material up to 2025. Its sellers say it can reason step by step and call a tool. The catalogue files it under chat. Its weights are published under APACHE-2.0, so you may run it on your own machine, or buy it from whoever serves it cheapest.
Every current figure
- Maker
- stepfun-ai
- Register
- model
- Takes
- text + image
- Returns
- text
- Context
- 65,536 tokens
- Longest answer
- 64,000 tokens
- Published
- July 2025
- Knowledge to
- 2025
- Licence
- apache-2.0
Known as 1 name
stepfun-ai/step3