Local LLM model page

Phi-3.5 MoE Instruct

Microsoft Phi-3.5 MoE: compact mixture-of-experts model with only ~6.6B active parameters. Strong reasoning and coding for local power users. MIT licensed.

Find the best model for my hardwareBrowse all 183 LLMs
Parameters
42B (6.6B active, MoE)
Minimum RAM
32 GB
Model size
24 GB
Quantization
Q4_K_M

Can Phi-3.5 MoE Instruct run locally?

Phi-3.5 MoE Instruct is best suited for power-user machines with 32 GB RAM. LocalClaw recommends Q4_K_M as the default quantization, with at least 32 GB RAM.

Search term for LM Studio or compatible runtimes: phi-3.5-moe-instruct

Hugging Face repository: microsoft/Phi-3.5-MoE-instruct

chatcodereasoningpowermoe

Strengths

  • Microsoft Phi-3.5 MoE: compact mixture-of-experts model with only ~6.6B active parameters. Strong reasoning and coding for local power users. MIT licensed.

Limitations

  • Performance depends heavily on quantization, RAM bandwidth and runtime support.

Best use cases

  • chat
  • code
  • reasoning
  • power
  • moe

Benchmarks

Speed: 6/10

Quality: 8/10

Coding: 8/10

Reasoning: 8/10

Technical details

Developer: phi

License: See model repository

Context window: Unknown tokens

Architecture: See model card

Released: 2025-12