# Step 3.7 Flash (slim)

> Step 3.7 Flash is StepFun's open-weight vision-language model, 198B parameters with about 11B active, with tool calling and a 256K context. Apache 2.0. Its 4-bit GGUF is 105 GB, just over what a 128 GB machine holds.

- Full: https://www.anchorterminal.com/tools/step-3-7-flash.md (~9,900 tokens) · this version ~1,630 tokens · JSON https://www.anchorterminal.com/tools/step-3-7-flash.json · canonical https://www.anchorterminal.com/tools/step-3-7-flash
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-10 · pricing https://www.anchorterminal.com/tools/step-3-7-flash#pricing · reviews https://www.anchorterminal.com/tools/step-3-7-flash#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**C · 59.8/100 · rank #93 of 160 open-weight models · #73 in Open-weight models · not agent-ready · confidence low** · same model listed separately: StepFun API, Model API (https://www.anchorterminal.com/tools/stepfun-api.min.md)


Assessment: A capable open agent model with a clean licence, every kind of build and four released runtimes. Two things hold it back: it's a little too big for a 128 GB machine at 4 bits, whatever the card says, and StepFun has gone quiet on the repository since June. No safety results, limitations or data description.

## Facts

- Kind: Open-weight model · vendor: StepFun · category: Open-weight models · legal entity: STEPFUN TECHNOLOGY (HK) LIMITED · provenance 58/100
- Packages: huggingface `stepfun-ai/Step-3.7-Flash`, huggingface `stepfun-ai/Step-3.7-Flash-GGUF`, huggingface `stepfun-ai/Step-3.7-Flash-FP8`, huggingface `stepfun-ai/Step-3.7-Flash-NVFP4`
- Auth: None · pricing: Free · x402: no · licence: Apache-2.0 (metadata and card; no LICENSE file)
- Benchmark probes: not run yet, so Performance is pending
- Surface graded: The open weights at huggingface.co/stepfun-ai/Step-3.7-Flash and the GGUF, FP8 and NVFP4 repositories beside it, read from the model card, the files and their metadata, the config, the chat template, the commit history and the runtimes' own code at their latest releases on 10 November 2026. StepFun's company facts come from our StepFun API listing. The model page on stepfun.com, the repository's Python files and the pre-built Docker images weren't read. Nothing was downloaded or run
- Model: 201,365,316,160 parameters in the files. The card says 198B: a 196B sparse mixture-of-experts language model with about 11B active per token, 45 layers, and a 1.8B vision encoder. Step3p7ForConditionalGeneration
- Files: 26 safetensors files in BF16 (402.7 GB), chat_template.jinja, config and tokenizer files and four Python modules. No pickle files. The GGUF repository has IQ3_XXS to BF16 with a projector and MTP files. Hugging Face lists a SHA-256 for the large files
- Licence: Apache 2.0 in the metadata of every repository and in the card's licence section. There is no LICENSE file
- Runtimes: The card documents vLLM and SGLang, each from a pre-built image, Transformers 5.0 or later with trust_remote_code for checking, and llama.cpp. Released vLLM 0.33.0, SGLang 0.5.23 and Transformers 5.22.2 have the step3p7 model, and llama.cpp b12242 has the architecture, with StepFun's own GGUF
- Context: 256K tokens per the card, and 262,144 positions in config.json. The tokenizer config says 131,072, and the card's NVFP4 example serves at 8,192
- Tools and reasoning: The card's vLLM and SGLang commands name the step3p5 tool-call parser, with auto tool choice on vLLM, and the step3p5 reasoning parser. The chat template takes tools and a reasoning_effort setting and keeps reasoning_content. The card describes tool orchestration results and doesn't show the setting in a request
- Hardware: The card says data-centre GPUs for production, and for local use machines with at least 128 GB of unified memory: NVIDIA DGX Station, AMD Ryzen AI Max+ 395 systems and Mac Studio or MacBook Pro. Its serve commands use 8 GPUs for BF16 and FP8 and 4 for NVFP4
- Training data: Not described on the card

- Scores: Capability pending, Reach 70, Risk 45, Agent interface 85, Documentation and upkeep 39 · total over the 7 assessed categories
- Why:  · Reach, Memory 5, the smallest build at 4 bits is StepFun's IQ4_XS GGUF, 104.99 GB with the 3.97 GB projector, and 108.96 GB × 1.2 = 130.8 GB, over… · Risk, File safety 15, safetensors and GGUF only, no .bin, .pt or .pth files (25). · Agent interface, Chat template 20, chat_template.jinja in the repository (5,723 characters). · Documentation and upkeep, Model card 14, Intended use, the architecture, the parameters, the context and the language (in the metadata) are on the card.
- Sources: 16, open questions: 5, both in the full twin
- Capabilities: model.open-weights
- JSON: https://www.anchorterminal.com/api/v1/tools/step-3-7-flash.json
- Its vendor verifies the listing with `verify_listing` or `POST https://www.anchorterminal.com/api/v1/verify` (the full twin says how); no effect on the grade.
- Support: https://platform.stepfun.ai/developer-community; more in the full twin

## Before you call it

1. On a 128 GB machine the 4-bit GGUF won't fit with a working context. Use StepFun's API or a provider, or a server
2. On vLLM or SGLang pass `--tool-call-parser step3p5 --reasoning-parser step3p5`
3. The card's commands pass trust-remote-code. Released Transformers 5.22.2 has the model built in if you'd rather not run the repository's code
4. For llama.cpp load the mmproj file from StepFun's GGUF repository for image input
5. StepFun's API has two hosts. Use the one where your key was issued, `api.stepfun.ai` or `api.stepfun.com`
6. The same model is on our StepFun API listing, with prices and terms

## Connect

```bash
pip install "sglang>=0.5.23"
sglang serve --model-path stepfun-ai/Step-3.7-Flash-FP8 --tp 8 --ep 4 --reasoning-parser step3p5 --tool-call-parser step3p5 --enable-multimodal --trust-remote-code
```

```bash
# SGLang and vLLM serve an OpenAI-compatible API. StepFun's own API is at https://api.stepfun.ai/v1
# (https://api.stepfun.com/v1 for keys issued in China)
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Qwen3-Embedding-8B | BB | 75.4 | model.open-weights | https://www.anchorterminal.com/tools/qwen3-embedding-8b.min.md |
| Gemma 4 31B IT | BB | 74.1 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-31b-it.min.md |
| Qwen3-Embedding-4B | BB | 73.6 | model.open-weights | https://www.anchorterminal.com/tools/qwen3-embedding-4b.min.md |
| Gemma 4 12B IT | BB | 72.3 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-12b-it.min.md |
| Gemma 4 26B A4B IT | BB | 72.2 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-26b-a4b-it.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)