stepfun-ai/step3
Model Card
Step3 is a multimodal reasoning model from StepFun built on a Mixture-of-Experts architecture with 321B total parameters (38B active), designed for cost-effective vision-language reasoning. It features innovative Multi-Matrix Factorization Attention (MFA) and Attention-FFN Disaggregation (AFD) for efficient inference across various hardware. The model excels at visual interpretation, complex reasoning, and is available via API or as open-source checkpoints on Hugging Face.
Context Window
N/A
tokens
Max Output
65,536
tokens
Input Cost
$0.57
per million tokens
Output Cost
$1.42
per million tokens
API Usage Example
Add StepFun: Step3 to your app with just a few lines of code.
No API keys, no backend, no configuration required.
<html>
<body>
<script src="https://js.puter.com/v2/"></script>
<script>
puter.ai.chat("Explain quantum computing in simple terms", {
model: "stepfun-ai/step3"
}).then(response => {
document.body.innerHTML = response.message.content;
});
</script>
</body>
</html>
Get started with Puter.js
Add StepFun: Step3 to your app without worrying about API keys or setup.
Read the Docs View Tutorials