StepFun: Step3

Other

Multimodal

Paid

Step3 is a cutting-edge multimodal reasoning model—built on a Mixture-of-Experts architecture with 321B total parameters and 38B active. It is designed end-to-end to minimize decoding costs while delivering top-tier performance in vision–language reasoning. Through the co-design of Multi-Matrix Factorization Attention (MFA) and Attention-FFN Disaggregation (AFD), Step3 maintains exceptional efficiency across both flagship and low-end accelerators.

Parameters

Context Window

65,536

tokens

Input Price

$0.57

per 1M tokens

Output Price

$1.42

per 1M tokens

Capabilities

Model capabilities and supported modalities

Performance

Reasoning

Excellent reasoning capabilities with strong logical analysis

Math

Coding

Knowledge

Modalities

Input Modalities

image,text

Output Modalities

text

LLM Price Calculator

Calculate the cost of using this model

Input Tokens (0.00000057/token)$0.000855

Output Tokens (0.00000142/token)$0.004260

Common Scenarios:

Input Cost:$0.000855

Output Cost:$0.004260

Total Cost:$0.005115

Estimated usage: 4,500 tokens

Monthly Cost Estimator

Based on different usage levels

Light Usage

$0.0199

~10 requests

Moderate Usage

$0.1990

~100 requests

Heavy Usage

$1.9900

~1000 requests

Enterprise

$19.9000

~10,000 requests

Note: Estimates based on current token count settings per request.

Last Updated: 1970/01/21