// StepFun Step · Text LLM
Step-3.5-Flash: VRAM & local deployment
196B total · 11B active (MoE) parameters. About 135 GB at 4-bit and 436 GB at BF16 with one 8K-token request.
released 2026-02-01licence: Commercial OKmemory figures are estimates
01Specs
| Publisher | StepFun · stepfun-ai |
|---|---|
| Family / series | StepFun Step · Step-3.5-Flash |
| Type | Text LLM (text-generation) |
| Parameters | 196B total · 11B active (MoE) Source: official model card“# Activated Params 11B · # Total Params (MoE) 196B” |
| Context length | 256K tokens (262,144)max_position_embeddings in config.json |
| Architecture | 45 layers (12 full-attention) · 36 sliding-window layers (512) · GQA · MoE |
| Released weights | BF16 · 399 GBdownload |
| Licence | Apache-2.0 · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed. |
| Released | 2026-02-01 (Hugging Face repo created) |
02How much VRAM it needs (estimate)
Total = weights + 10% + 0.5GB overhead + KV cache for one request (FP16 KV). More concurrent users multiply the KV column.
| Precision | Weights | Total @ 8K | Total @ 32K | Total @ 128K | Smallest GPU | Mac |
|---|---|---|---|---|---|---|
| BF16 / FP16 (original) | 392 GB | 436 GB | 445 GB | 484 GB | 8 × H100 (640GB) | — |
| FP8 / INT8 / Q8_0 | 208 GB | 233 GB | 243 GB | 282 GB | 4 × RTX PRO 6000 (384GB) | Mac 512GB |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 119 GB | 135 GB | 145 GB | 183 GB | 4 × RTX PRO 6000 (384GB) | Mac 256GB |
KV cache per request: 8K 3.8 GB · 32K 13.5 GB · 128K 52.1 GB
Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →
03How to run it
Only runtimes the publisher or the runtime’s official library names for this model.
- vLLM — mentioned in the official model card
- SGLang — mentioned in the official model card
- llama.cpp — mentioned in the official model card
- Transformers — mentioned in the official model card
04Official repositories & quantised builds
4 official repo(s) from stepfun-ai.
- Step-3.5-Flash main
- Step-3.5-Flash-FP8 FP8
- Step-3.5-Flash-Base Base
- Step-3.5-Flash-GGUF-Q8_0 GGUF
05Other StepFun Step models
All 12 →| Model | Params | 4-bit total | Released |
|---|---|---|---|
| Step-5-Preview | 600B / 27B | 412 GB | 2026-09-20 |
| Step-3.7-Flash | 198B / 11B | 136 GB | 2026-05-23 |
| Step-Audio-R1.1 | 33.5B | 25.0 GB | 2026-01-14 |
| Step3-VL-10B | 10.2B | 8.5 GB | 2026-01-13 |
| Step-Audio-EditX | 3.53B | 6.1 GB | 2025-10-29 |
| Step-Audio-2-mini | 8.31B | 6.5 GB | 2025-08-28 |
| step3 | 321B / 38B | 247 GB | 2025-07-28 |
| Step-Audio-AQAA | 136.7B | 127 GB | 2025-06-11 |
06Sources
Data checked 2026-03-17. Parameters, context and licence are copied from the official repository; memory figures are our estimates.