lkevincc0/Step-3.5-Flash-REAP-128B-A11B

Step-3.5-128B-A11B | Technical Specifications

Attribute	Detail
Base Model	Step-3.5-Flash
Architecture	Sparse Mixture-of-Experts (SMoE)
Model Type	Causal Language Model
Total Parameters	128B
Active Parameters	11B (per token)
Compression Ratio	40% (Expert Pruning)
Pruning Strategy	Pruning observation (REAP)
Calibration Set	lkevincc0/glm47-math-code-calibration-1024

---

Local Deployment

Step 3.5 Flash is optimized for local inference and supports industry-standard backends including vLLM, SGLang, Hugging Face Transformers, and llama.cpp.

vLLM

We recommend using the latest nightly build of vLLM.

1. Install vLLM

# via Docker
docker pull vllm/vllm-openai:nightly

# or via pip (nightly wheels)
pip install -U vllm --pre \
  --index-url https://pypi.org/simple \
  --extra-index-url https://wheels.vllm.ai/nightly

2. Launch the server

Note: Full MTP3 support is not yet available in vLLM. We are actively working on a Pull Request to integrate this feature, which is expected to significantly enhance decoding performance.

• For FP8 model

vllm serve <MODEL_PATH_OR_HF_ID> \
  --served-model-name step3p5-flash \
  --tensor-parallel-size 8 \
  --enable-expert-parallel \
  --disable-cascade-attn \
  --reasoning-parser step3p5 \
  --enable-auto-tool-choice \
  --tool-call-parser step3p5 \
  --hf-overrides '{"num_nextn_predict_layers": 1}' \
  --speculative_config '{"method": "step3p5_mtp", "num_speculative_tokens": 1}' \
  --trust-remote-code \
  --quantization fp8

• For BF16 model

vllm serve <MODEL_PATH_OR_HF_ID> \
  --served-model-name step3p5-flash \
  --tensor-parallel-size 8 \
  --enable-expert-parallel \
  --disable-cascade-attn \
  --reasoning-parser step3p5 \
  --enable-auto-tool-choice \
  --tool-call-parser step3p5 \
  --hf-overrides '{"num_nextn_predict_layers": 1}' \
  --speculative_config '{"method": "step3p5_mtp", "num_speculative_tokens": 1}' \
  --trust-remote-code

You can also refer to the Step-3.5-Flash recipe.

---

SGLang

1. Install SGLang

# via Docker
docker pull lmsysorg/sglang:dev-pr-18084

# or from source (pip)
pip install "sglang[all] @ git+https://github.com/sgl-project/sglang.git"

2. Launch the server

• For BF16 model

sglang serve --model-path <MODEL_PATH_OR_HF_ID> \
  --served-model-name step3p5-flash \
  --tp-size 8 \
  --tool-call-parser step3p5 \
  --reasoning-parser step3p5 \
  --speculative-algorithm EAGLE \
  --speculative-num-steps 3 \
  --speculative-eagle-topk 1 \
  --speculative-num-draft-tokens 4 \
  --enable-multi-layer-eagle \
  --host 0.0.0.0 \
  --port 8000

• For FP8 model

sglang serve --model-path <MODEL_PATH_OR_HF_ID> \
  --served-model-name step3p5-flash \
  --tp-size 8 \
  --ep-size 8 \
  --tool-call-parser step3p5 \
  --reasoning-parser step3p5 \
  --speculative-algorithm EAGLE \
  --speculative-num-steps 3 \
  --speculative-eagle-topk 1 \
  --speculative-num-draft-tokens 4 \
  --enable-multi-layer-eagle \
  --host 0.0.0.0 \
  --port 8000

---

Reference

• Hugging Face: stepfun-ai/Step-3.5-Flash
• Optimization Tech: CerebrasResearch/reap