# Deploy Step-5-Preview on dedicated GPUs

Step-5-Preview by StepFun, deployable on sference as a dedicated deployment: reserved GPU capacity, private endpoint, European GPUs, billed per GPU-hour on monthly commitments. Request it at https://sference.com/models/stepfun-ai/Step-5-Preview.

## Specs

- Hugging Face: https://huggingface.co/stepfun-ai/Step-5-Preview
- Category: Chat & reasoning
- Parameters: 604B
- Architecture: MoE · 352 experts, 8 active
- Context: 1M tokens
- Published precision: FP8 + BF16
- Checkpoint size: 650 GB
- Minimum footprint: 2 nodes × 8 GPUs (80 GB) or 8 GPUs (141 GB) to load the weights; more for long contexts and concurrency
- License: Custom license (Review terms) — https://huggingface.co/stepfun-ai/Step-5-Preview/blob/main/LICENSE.txt

## 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.

Facts from Hugging Face, synced 2026-10-27. All deployable models: https://sference.com/dedicated.md