Step3-VL-10B

Checkpointstepfun-ai/Step3-VL-10BModel card indexedModel card

Support summary

Step3-VL-10B is a checkpoint published by StepFun, built on the StepVLForConditionalGeneration architecture. The newest indexed releases vLLM 0.28.0 and SGLang 0.5.19 both document support.

vLLM documents support in 20 of 20 indexed releases, from 0.17.0 to 0.28.0. SGLang documents support in 13 of 13 indexed releases, from 0.5.10 to 0.5.19.

Documented features
Reasoning parsers
Configured context
65,536 tokens
License
apache-2.0

Hardware, quantization and features count only statements about this checkpoint; engine-wide capabilities are listed with each release.

Hugging Face model card
Publisher
StepFun
Task
image text to text
License
apache-2.0
Architecture
StepVLForConditionalGeneration
Weight tensor elements
10.2B
Tensor types
BF16
Configured context
65,536 tokens
Base models
stepfun-ai/Step3-VL-10B-Base
Weight access
Public
Repository created
2026-01-13
Source 1: stepfun-ai/Step3-VL-10B · publisher model cardSource 2: stepfun-ai/Step3-VL-10B · Hugging Face metadataSource 3: stepfun-ai/Step3-VL-10B · publisher configurationReviewed
Model notes and sources

Engine support

3 engines

Engine support for Step3-VL-10B
EngineReleaseCheckpoint supportEvidence scopeLast known supported
vLLM0.28.0Newest indexedCheckpoint evidenceSource: vLLM v0.28.0 · docs/models/supported_models.mdReviewed 0.28.0This release
Features, dependencies and evidence for vLLM 0.28.0
SGLang0.5.19Newest indexedCheckpoint evidenceSource: SGLang v0.5.19 · docs/docs/supported-models/multimodal_language_models.mdxReviewed 0.5.19This release
Features, dependencies and evidence for SGLang 0.5.19
TensorRT-LLM1.2.1Newest indexedNo matching evidence curatedNo supported release indexed
Features, dependencies and evidence for TensorRT-LLM 1.2.1

Recipes and cookbooks