Step3-VL-10B
Official · BF16 · safetensors
Runtime certified: Transformers
| Precision / quantization | Format | Bits | Publisher trust | Deployment state |
|---|---|---|---|---|
| BF16 | safetensors | 16 | Official | Runtime certified |
StepFun · Vision · Step3-VL
Step3-VL 10B: Specs, GPU requirements, hosting & deployment
Step3-VL 10B is a current Vision release in the StepFun family, certified from its immutable official source revision.
Total parameters10.0B
Active parametersDense model
Context window65,536 tokens
Release statusCurrent · Current
ArchitectureDense Transformer
ModalitiesText, Image, Video
LicenseApache License 2.0
Deployment stateDeployable with certified configuration
Commercial stateCommercial Use Allowed
Release date2026-01-13
Architecture, strengths and use cases
Step3-VL 10B is an official self-hostable StepFun release in the StepFun Vision track. It uses Dense Transformer, carries 10B, supports a verified context window of 65,536 tokens, and exposes only source-certified checkpoints and runtime combinations. Its release, license and deployment facts are tied to the official repository revision rather than inferred from the family name.
Step3-VL 10B uses Dense Transformer with 10B. The stored architecture profile records 64 layers, hidden width 8192, 64 attention heads and a 65,536 token context ceiling where the official configuration publishes those fields.
Verified Dense Transformer architecture
65,536 token certified context
Official BF16, FP8 checkpoint coverage
Private LLM inference
Agentic workflows
Long-context document processing
Vision-language analysis
Certified checkpoints and quantizations
Only exact, source-certified variants attached to this release are shown.
Step3-VL-10B-FP8
Official · FP8 · safetensors
Runtime certified: Transformers
| Precision / quantization | Format | Bits | Publisher trust | Deployment state |
|---|---|---|---|---|
| FP8 | safetensors | 8 | Official | Runtime certified |
Recommended hardware
Use only the listed transformers runtime configuration with the exact checkpoint precision. The recommendation engine checks weight memory, KV cache, runtime overhead, safety headroom, GPU architecture and topology before ranking rentable and reference hardware.
Estimated weight memory19.19 GB
KV cache2 GB
Runtime overhead8.53 GB
Total with safety margin36.85 GB
Recommended best value
1× NVIDIA H200 SXM 141GB
141 GB total VRAM · nvswitch
104.15 GB headroom · Limited capacity
USD 2,345 / monthLowest-cost viable
2× NVIDIA L4 24GB
48 GB total VRAM · pcie
11.15 GB headroom · Available
USD 500 / monthHighest performance
1× NVIDIA H200 SXM 141GB
141 GB total VRAM · nvswitch
104.15 GB headroom · Limited capacity
USD 2,345 / monthLicense and self-hosting considerations
The full 10B resident weight set must fit across the selected GPUs. Context and concurrency add KV-cache memory; MoE active parameters do not replace resident-weight memory.
Step3-VL 10B is published under Apache License 2.0. The release-scoped official license link and verification date are retained with the catalogue record.
Commercial Use Allowed. The official license permits commercial use; its obligations still apply.
Release timeline
stepfun/stepfun/step3-vl-10b
Source details
Source revision: 6834d893eb35cd0c0af8e24c76c9de39b786c778637069212687e46a5b135c0a
Frequently asked questions
How much VRAM does Step3-VL 10B require?
The exact estimate depends on checkpoint precision, context and concurrency; the page recommendation uses the certified memory engine and compatible runtime configurations.
Which runtimes support Step3-VL 10B?
Certified upstream runtime evidence is available for transformers.
Can Step3-VL 10B be used commercially?
The stored commercial classification is Commercial Use Allowed. Consult the official Apache License 2.0 terms for the final obligations.
Official references
Last verified: 2026-08-26