Models · StepFun · Chat & reasoning

Deploy Step-5-Preview

stepfun-ai/Step-5-PreviewdedicatedHugging Face ↗

Step-5-Preview by StepFun on GPU capacity reserved for you: a private OpenAI-compatible endpoint, every request processed on European GPUs and logged in your audit trail, billed per GPU-hour on a monthly commitment.

Request Step-5-PreviewHow dedicated deployments work →
Maker
StepFun
Params
604B
Context
1M
Architecture
MoE · 352 experts, 8 active
License
Custom license
Released
2026-09-20

Footprint

Checkpoint
650 GB
FP8 + BF16
80 GB-class GPUs
2 nodes × 8 GPUs
minimum to load
141 GB-class GPUs
8 GPUs
minimum to load

Minimum to load the published checkpoint with a small KV cache. Long contexts and high concurrency need more; quantizing a 16-bit checkpoint needs less. You don't pick hardware on sference: you choose a performance tier and we size and run the deployment.

License

Custom licenseReview terms

Custom license from the model maker. Review the terms on the model page before production use. Read the license. Not legal advice.

FAQ

Can I run Step-5-Preview on a dedicated deployment in Europe?

Yes. sference deploys Step-5-Preview on GPU capacity reserved for you, behind a private endpoint, with every request processed on European GPUs and logged in your audit trail. Tell us your expected volume and latency targets with the form on this page and we come back with a sizing and a quote.

How many GPUs does Step-5-Preview need?

The published checkpoint is about 650 GB (FP8 + BF16), so loading it takes at least 2 nodes × 8 GPUs with 80 GB of memory each, or 8 GPUs with 141 GB each. Long contexts and high concurrency need more for the KV cache; quantization needs less. You don't pick hardware on sference: you choose a performance tier and we size the deployment.

Is Step-5-Preview licensed for commercial use?

Step-5-Preview is published under the Custom license license. Custom license from the model maker. Review the terms on the model page before production use.

How is a dedicated deployment priced?

Per GPU-hour on a monthly commitment, sized to your throughput and latency targets.

Related models

StepFun
Step-3.5-Flash
199B · 256K context
StepFun
Step-3.7-Flash
201B · 256K context
StepFun
Step3-VL-10B
10B · 64K context
DeepSeekserverless
DeepSeek-V4.1-Flash
763B · 1M context
Z.aiserverless
GLM-5.3-Flash
321B · 1M context
Z.aiserverless
GLM-5.3
753B · 1M context

Request Step-5-Preview

Tell us roughly how much traffic you expect. A fine-tune of Step-5-Preview? Mention it in the notes.

  1. 01
    We size it

    A performance tier, the capacity behind it, and a quote for your volume.

  2. 02
    We benchmark it

    Bring one workload with a sample of real traffic; we come back with latency and throughput numbers.

  3. 03
    It goes live

    We provision the deployment. You call it with the same API and model id as before.

No commitment. Prefer email? [email protected]

Specs from Hugging Face, synced 2026-10-27. Machine-readable version: https://sference.com/models/stepfun-ai/Step-5-Preview.md