StepFun 5 Preview GGUF

GGUF quants of StepFun's 604B-parameter flagship MoE model.

Playable

Description

Quantized versions (IQ1_S, Q2_K, Q3_K_M) of the Step-5-Preview sparse MoE model. The project includes technical fixes for rope frequency base metadata and provides companion files for parts of the architecture, like the perception encoder and MTP layers, not yet supported by llama.cpp.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.

Details

Why this is interesting: Technical feat of quantizing a 600B+ parameter model into 1-bit and 2-bit formats for local execution on high-end…
Attributed to: Cruz@vic305
Built with: Step 5 Preview Redwood AI
Links: huggingface.co
View tweet
Tags: #gguf #quantization #moe #llm

Discovered: X
First posted: X · Oct 15, 2026
Listed: Added 2d ago
Build: Strong
Models: Strong

StepFun 5 Preview GGUF

Description

How it grew

StepFun 5 Preview GGUF

Description

How it grew