Skip to content
#16A+Aug 20, 2026Evidence B

Step 3.7 Flash

StepFun · 196B total / 11B active (MoE) + 1.8B ViT

A low-cost tier for high-volume multimodal work

An August 2026 196B-A11B multimodal tier with a 1.8B ViT, open weights, 256K context and ¥2/1M output tokens. Perception 92, documents 92 and agents 91 give a composite of 89.2; generation 78 and ecosystem 84 keep it out of A+.

TextImageVideoNative

Last updated Oct 21, 2026Editorial scoring

Bottom line

Think of Step 3.7 Flash as an open-weight fast tier for documents and agents: a fraction of flagship pricing, adequate documents and tool use, and 256K context for a whole manual. Do not use it for creative generation, and have legal review the community licence.

Eight-dimension breakdown

editorial mapping · 0–100
Perception92Document92Chart QA89Video88Agent91Generation78Chinese88Ecosystem848D
  • Step 3.7 Flash89.2
Visual perception & grounding20%
92
Document parsing & Chinese OCR18%
92
Chart & document reasoning14%
89
Video & long-horizon reasoning14%
88
GUI agents & tool use14%
91
Generation & editing6%
78
Chinese context & culture6%
88
Ecosystem, cost & availability8%
84

Weights: perception 20 / documents 18 / charts 14 / video 14 / agents 14 / generation 6 / Chinese 6 / ecosystem 8. Definitions and sources on the methodology page.

Public benchmarks

as reported, not converted
Benchmark / metricValueNote
参数规模196B-A11BPlus a 1.8B ViT
上下文256KOfficial spec
输出价格¥2Per 1M tokens

Public scores come from different versions and prompts and are not interchangeable with our composite; they are one evidence input.

Scenario mapping

five stress tests, one scale
ScenarioScoreObservation
Chinese receipt OCR92 · Stable on clean receipts
Chart data extraction89 · Reliable on standard charts
30-min video QA88 · Above-average event localization
Mobile GUI control90 · Well-formed tool calls
Chinese cultural context88 · Good official register; average on memes

Strengths and limitations

Strengths

  • ¥2/1M output tokens at 196B-A11B: a cost-friendly tier for high-volume documents and agents.
  • Documents 92, agents 91 and 256K context are a rare combination at this price.
  • Open weights let it pair with Step 5 Preview in a triage-plus-review relay.

Limitations

  • Generation 78: no creative generation or complex image editing.
  • Ecosystem 84: SDK and hosted-service maturity trail Qwen and Zhipu.
  • The StepFun community licence is not OSI open source; commercial terms need review.

Price and real cost

List price

¥2/M tokens

≈ ¥2 / 1M output tokens; weights free. For 1,000 images at 1,024px, ~1,100 image tokens each and 300 output tokens, the monthly cost is about ¥1.2 (assuming input at 25% of output price).

Use your volume

About Step 3.7 Flash

Is Step 3.7 Flash open source?

Licence: Open weights (StepFun community licence). Weights are available, but check the exact terms and filing requirements before commercial use.

How much does Step 3.7 Flash cost?

≈ ¥2 / 1M output tokens; weights free. It ranks #16 with a composite of 89.2. Real bills depend on resolution, tiling, caching and volume discounts.

What is Step 3.7 Flash best for?

High-volume receipt and document batches; Cost-sensitive agent orchestration; Video content triage。

Why does Step 3.7 Flash score 89.2?

The composite weights eight dimensions: perception 20%, documents 18%, charts 14%, video 14%, agents 14%, generation 6%, Chinese 6%, ecosystem 8%. Its strongest dimension is Visual perception & grounding; its weakest is Generation & editing.

Personal ranking · Not an official score

Every ranking, score and comparison on this page is an independent personal ranking, not an official score, and represents no vendor, institution or regulator. Scores are a subjective mapping of public evidence under our own weights — change the weights and the order changes. Disagree, or spotted a factual error? Contact the site owner: jizhi2024@foxmail.com — we check each report and log corrections in the changelog.