StepFun has been focusing on MIMO recently, rolling out a free trial with no barriers. But until my MIMO plan expires, I’d better stick with MIMO. If any friends have experience using StepFun, feel free to share it in the comments. https://platform.stepfun.com/step-plan#plan-list…

StepFun Open Platform

Source: https://platform.stepfun.com/step-plan

Massive Tokens Giveaway! Register now and get a 15-day free trial, available to both new and existing users.

Use them up within 15 days, and get another 15-day trial! For every friend you invite to register, you’ll earn an extra 15-day free trial, up to a maximum of 90 days!

Offer ends: July 18, 2026. 26 days 4 hours remaining.

Plans

New Step Plan is live — unlimited quotas plus multiple perks! Come try it now.

Flash Mini - Simple & affordable, suited for beginners exploring AI
New users: free trial for 15 days
Click to claim

Flash Plus - For everyday AI productivity
Subscribe

Flash Pro - For frequent heavy AI users
Subscribe

Flash Max - For professional power users
Subscribe

Unified Credit system — usage is calculated by token consumption. Supports plan upgrades and usage tracking.

Developer Reviews

戴以沫
The first-generation gelab was a purely visual small model. It had low recognition accuracy in complex scenarios like semi-transparent pop-ups, ad pop-ups, and CAPTCHAs, often hallucinated during multi-step operations, and frequently missed digits in number input. After upgrading to the multimodal large model, adaptability to complex scenarios improved dramatically — the pass rate for batch multi-step operations reached 96%, significantly enhancing end-to-end delivery stability. The upgrade to Step 3.7 Flash is not just about optimizing visual capabilities; it achieves a core qualitative leap from “reading screen content” to “stably running the full workflow.” This full-chain stability is the key foundation for Agent-to-Agent (A2A) collaboration: stable individual agent capabilities are a prerequisite for multi-agent collaboration. This breakthrough greatly reduces the cost of building and debugging future A2A systems.

语码Cace
As a long-time music player, I first learned about StepFun from the first generation of ACESTEP. During the hackathon, to demonstrate API response speed, we used Step 3.7 Flash — it was incredibly solid, never faltering during live demos. The new Realtime TTS model is the most impressive; it can assist me in alternating interpretation practice, and its response speed is even faster than a real teacher!

Cogito
Foresee AI Ca’s CTO
I have to give a huge shout-out to Step 3.7 Flash and StepAudio — the speed is incredible, the fastest I’ve ever seen! There’s none of that anxious waiting while watching the screen spin. Especially Step 3.7 Flash — it’s the ultimate “cost-performance king,” cheap and generous. I’ve fully migrated my locally built stick figure Agent OS to this architecture — it’s truly the best partner. Whether running long tasks or everyday agents, these two are the perfect “token food” combo — feeding fast and full, maxing out productivity. Absolutely top-notch!

沫音
I’ve been a long-time user of the Step series, from Step 3.5 to 3.7. I can really feel the rapid improvement in model intelligence, and the Step series has always maintained very high response speed. Especially with 3.7, you almost get the feeling this model has no “filter” (lol). Also, when writing tutorials, Step 3.7 Flash’s language skills are really good — I only need minor tweaks to meet my needs. The experience is top-notch!

筱可
After version 3.7, Step’s models have significantly improved in programming capabilities, and they also have decent visual capabilities — supporting video, voice, images, and text multimodal capabilities. Very impressive!

background

Ready to start your AI journey?

Join over 10,000+ developers and creators, and experience AI services with fixed costs and unlimited innovation.

Low-cost startup · Cancel anytime · Extra discounts for enterprise users