Skip to content

Catalogmodels-api

StepFun API

situationalBlueshift — gaining momentum

API for Step-2/Step-3 family of frontier MoE models from Shanghai.

Visit platform.stepfun.com →Add to kitOpen in Noemium

ledger entry · observed by @whysanesanders

Step 3.7 Flash is cheap, open weights, and scores well on coding and math benchmarks. Step-2 is a capable closed model at mid-tier prices. Brand recognition and docs outside China are weak, and access often runs through resellers.

Strengths

  • Step 3.7 Flash is cheap, open weights, and strong on coding and math.
  • Step-2 is a capable closed model at mid-tier prices.
  • The family covers text, vision, and reasoning variants.

Known limitations

  • The hosted API is proprietary; Step 3.5 Flash and Step 3.7 Flash weights are separate open-weight releases.
  • StepFun runs separate China (platform.stepfun.com) and global (platform.stepfun.ai) platforms with non-interchangeable API keys.
  • Lower Western brand recognition; docs mostly Chinese.
  • Mixed access outside China; often reached via resellers.
  • Step-2 is closed, unlike Step 3.5 Flash and Step 3.7 Flash.
  • Smaller model family than Qwen or Tencent.

Use it when

  • Cost-sensitive coding and math workloads that fit the Flash SKUs.
  • Teams that want Step-2's closed-model quality without premium pricing.

Skip it when

  • You need strong Western brand recognition and English-first docs.
  • You want one API key that works across China and global platforms.
  • You need a model family as broad as Qwen or Tencent.

Buy-check

  • ·Price — paid, Step 3.7 Flash ~$0.20/$1.15 per 1M; Step 3.5 Flash open-source $0.10/$0.30; Step-2 ~$0.83/$2.50; Step-2-mini ~$0.14/$0.42; Step-1.5V ~$0.42/$1.25; Step-R ~$1.11/$3.33 as of 2026-08-24. Open the pricing page, not a screenshot.
  • ·Free tier — Free tier exists — burn it first.
  • ·Cancel fallout — What happens to your data/projects on cancel — check the ToS, not the FAQ.
  • ·Seat math — Per-seat pricing multiplies quietly. Count seats before the annual plan.
  • ·Who skips it — Also skipped when: You need strong Western brand recognition and English-first docs.; You want one API key that works across China and global platforms..

Price trail — 1 revisionchangelog →

  • 2026-W34Step-2 ~$0.83/$2.50 per 1M; Step-2-mini ~$0.14/$0.42;…Step 3.7 Flash ~$0.20/$1.15 per 1M; Step 3.5 Flash op…

Facts

pricing
paid1
price note
Step 3.7 Flash ~$0.20/$1.15 per 1M; Step 3.5 Flash open-source $0.10/$0.30; Step-2 ~$0.83/$2.50; Step-2-mini ~$0.14/$0.42; Step-1.5V ~$0.42/$1.25; Step-R ~$1.11/$3.33
free tier
yes — paid to keep using
open source
no
api
yes
self-host
no
category
models-api
evidence
source-verified
Models
locked

Sources

  1. 1platform.stepfun.ai[snapshot]
  2. 2github.com/stepfun-ai/Step-3.7-Flash[snapshot]
  3. 3apirank.vip/zh/tutorials/stepfun-api-review/[snapshot]
  4. 4tokenmix.ai/blog/step-3-5-flash-review-stepfun-2026[snapshot]

No source, no number.

Edit this page on GitHub

Similar joball alternatives →

toolverdictpriceossself-host
StepFun APIsituationalpaidnono
Google AI Studio / Gemini APIshipfreemiumnono
Anthropic APIshippaidnono
Cerebrasshipfreemiumnono

Facts from the catalog files — not scores you can buy. Quality and speed stay in the briefing, not in a fake 1–5 grid.

Google AI Studio / Gemini API
models-apiship

Google AI Studio / Gemini API

Gemini API with a real free tier — huge context, multimodal, generous limits.

APIFREE TIER

freemiumfree tier (text models); paid from $0.75/1M input (Gemini 3.7 Flash promo till 2027-01)

Blueshift — gaining momentum
Anthropic API
models-apiship

Anthropic API

Claude models direct — the practitioner's pick for coding and long-form text.

API

paidpay-as-you-go, from $1/1M input (Haiku 4.5); small free credits for new users

Blueshift — gaining momentum
Cerebras
models-apiship

Cerebras

Ultra-fast inference on wafer-scale AI accelerators with competitive API pricing.

APIFREE TIER

freemiumfree trial ($5 credits); self-serve Developer from $10 deposit; Enterprise custom

Blueshift — gaining momentum