llmrun

Step 3 Models — Hardware Requirements

6 Step 3 models from StepFun and the community, from the smallest that runs in 85.3 GB of VRAM at Q4 up to 321.0B parameters. Every row links to full quantization tables and GPU compatibility.

All Step 3 Models by Size

ModelParamsContext
Step 3.5 Flash199.4B262K
Step 3.7 Flash201.4B262K
Step3321.0B66K
Community variants and fine-tunes (3)
ModelParamsContext
Step 3.5 Flash REAP 121B A11B121.0B262K
Step 3.5 Flash Base197.8B33K
Step 3.5 Flash Base Midtrain197.8B131K

Frequently Asked Questions

How much VRAM do I need to run a Step 3 model?
The smallest Step 3 model, Step 3.5 Flash REAP 121B A11B, runs from 85.3 GB of VRAM at Q4_K_M (smaller low-bit quantizations start at 56.5 GB with a noticeable quality loss). Larger family members need proportionally more — see the table above for every model.
Which Step 3 models can I run on a 16 GB GPU?
No Step 3 model currently fits in 16 GB of VRAM — the family starts at 85.3 GB.
What is the most popular Step 3 model to run locally?
Step 3.7 Flash is the most downloaded Step 3 model in local-friendly quantized formats. It runs from 139.2 GB of VRAM at Q4_K_M.