πŸ“‹ Model Description

base_model:
  • stepfun-ai/Step-3.5-Flash

MoE-style quants of Step-3.5-Flash:

  • conditional experts quanted to varying degrees (IQ3S, IQ4S, Q4K, Q5K, ...)
  • rest of the model (attention, etc.) remains in Q8_0.

WIP, more quants + KLD measurements to come.

πŸ“‚ GGUF File List

No GGUF files available