What is StepFun
StepFun provides the Step family of multimodal reasoning, vision, speech, image and audio models through its Chinese-language developer platform.
Current API pricing
As of September 13, 2026, StepFun lists Step 3.7 Flash at CN¥1.35 per million uncached input tokens, CN¥0.27 cached input and CN¥8.10 output. Step 3.5 Flash and its 2603 snapshot each cost CN¥0.70 input, CN¥0.14 cached input and CN¥2.10 output. Step 1o Turbo Vision costs CN¥2.50 input, CN¥0.50 cached input and CN¥8 output.
Token-priced speech models include StepAudio 2.5 Realtime at CN¥10 input, CN¥2 cached input and CN¥70 output per million tokens; StepAudio 2.5 Chat at CN¥10, CN¥2 and CN¥25; Step 1o Audio at CN¥25, CN¥5 and CN¥60; Step Audio 2 at CN¥10, CN¥2 and CN¥70; and Step Audio R1.5 at CN¥10, CN¥2 and CN¥105.
Other media rates
Text-to-speech costs CN¥5.80 per 10,000 characters for StepAudio 2.5 TTS, CN¥2.80 for Step TTS 2 and CN¥0.90 for Step TTS Mini. Voice cloning costs CN¥9.90 per voice. Speech recognition ranges from CN¥0.15 per hour for StepAudio 2.5 ASR to CN¥2.60 for Step ASR 1.1 Stream. Step 2X Large costs CN¥0.10 per image and Step Image Edit 2 costs CN¥0.02 per image; both are scheduled to close on October 10, 2026.
These prices are published in Chinese yuan. They are not converted into US dollars for the API Cost sheet because the site does not currently store source currency and a conversion could mislabel the vendor's rate.
Official sources: StepFun pricing and rate limits and StepFun developer documentation.