stepfun Open weights Vision Reasoning

Step 3.7 Flash pricing & benchmarks

Released May 28, 2026 · served by 3 providers · ranked #54 by intelligence in our index

Input / 1M tokens$0.200
Output / 1M tokens$1.15
Context window262K
Cached input$0.040
Intelligence index30.3
Coding index39.6
Agentic index21.5
1M in + 300K out$0.55

What Step 3.7 Flash is

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...

Providers and prices for Step 3.7 Flash

The same model costs different amounts depending on who serves it. Prices are per 1M tokens, cheapest first. Uptime and throughput are OpenRouter's rolling measurements, not vendor claims.

ProviderInputOutputContext QuantUptime 24hThroughput
StepFuncheapest $0.200 $1.15 256K fp8 99.3%
DeepInfra $0.200 $1.15 262K 99.8%
Novita $0.200 $1.15 262K fp8 97.4%

Step 3.7 Flash benchmark results

Design Arena head-to-head results, as reported through the OpenRouter model API.

CategoryArenaRankEloWin rate
asciiart models #23 1193 46.7%
dataviz models #46 1203 45.9%
gamedev models #46 1204 41.9%
codecategories models #48 1203 44.9%
uicomponent models #48 1202 43.7%
website models #49 1212 46.3%
svg models #50 1122 39.3%
3d models #55 1178 42%

Cheaper models in the same class

Models scoring within 4 points of Step 3.7 Flash on the Intelligence Index, but with a lower output price.

FAQ

How much does the Step 3.7 Flash API cost?

$0.200 per 1M input tokens and $1.15 per 1M output tokens. A typical workload of 1M input + 300K output tokens costs about $0.55.

Is Step 3.7 Flash free?

No. It is a paid model starting at $0.200 per 1M input tokens, though some providers offer trial credits.

Which provider is cheapest for Step 3.7 Flash?

StepFun at $0.200 input / $1.15 output per 1M tokens (fp8 quantization).

Can I self-host Step 3.7 Flash?

Yes — weights are published as stepfun-ai/Step-3.7-Flash on Hugging Face, so you can run it on your own hardware.