---
id: "stepfun/step-3.7-flash"
also: ["stepfun/step-3.7-flash-20260528", "stepfun-ai/Step-3.7-Flash", "hf/stepfun-ai/Step-3.7-Flash", "requesty/stepfun/step-3.7-flash", "vercel/stepfun/step-3.7-flash"]
card_usd_per_m: "0.2 in, 1.15 out, at stepfun/fp8"
cheapest_list_usd_per_m: "0.16 in, 0.92 out, at deepinfra on openrouter"
cheapest_effective_usd_per_m: "0.0456 in, 0.919 out, at deepinfra on openrouter"
context: 262144
max_output: 230400
input: ["text", "image", "video"]
output: ["text"]
released: "2026-05-28"
reasoning: ["high", "medium", "low"]
api: ["/api/v1/chat/completions", "/api/v1/completions", "/api/v1/messages", "/api/v1/responses"]
params: ["frequency_penalty", "include_reasoning", "logit_bias", "logprobs", "max_tokens", "min_p", "presence_penalty", "reasoning", "reasoning_effort", "repetition_penalty", "response_format", "seed", "stop", "structured_outputs", "temperature", "tool_choice", "tools", "top_k", "top_logprobs", "top_p"]
endpoints: 7
gateways: "openrouter (3), hf (2), requesty (1), vercel (1)"
no_training: 4
trains: 1
retains_prompts: 2
coding: 39.6000
---
# StepFun: Step 3.7 Flash

Step 3.7 Flash is StepFun's latest high-efficiency multimodal Mixture-of-Experts model. It pairs a 196B-parameter language backbone with a vision encoder for native image and video understanding, activating roughly 11B parameters...