Local LLM model page

Qwen 3 MoE (30B/3B active)

Efficient MoE model with only 3B active params. Fast inference at large model quality. Hybrid thinking mode. Apache 2.0.

Find the best model for my hardwareBrowse all 183 LLMs
Parameters
30B (3B active)
Minimum RAM
24 GB
Model size
18 GB
Quantization
Q4_K_M

Can Qwen 3 MoE (30B/3B active) run locally?

Qwen 3 MoE (30B/3B active) is best suited for power-user machines with 32 GB RAM. LocalClaw recommends Q4_K_M as the default quantization, with at least 24 GB RAM.

Search term for LM Studio or compatible runtimes: qwen3-30b-a3b

Hugging Face repository: lmstudio-community/Qwen3-30B-A3B-GGUF

chatcodereasoningpowerspeed

Strengths

  • Only 3B active params — blazing fast
  • MoE efficiency at 30B quality level
  • Hybrid thinking mode
  • Apache 2.0
  • Great balance of speed and quality

Limitations

  • MoE architecture needs more disk space
  • Not as good as dense 32B for pure quality
  • Needs 24GB RAM despite efficient inference

Best use cases

  • Fast local inference
  • Real-time chat applications
  • Code completion
  • Speed-sensitive deployments
  • Multi-user serving

Benchmarks

Speed: 8/10

Quality: 8/10

Coding: 8/10

Reasoning: 8/10

Technical details

Developer: Alibaba Cloud (Qwen Team)

License: Apache 2.0

Context window: 131,072 tokens

Architecture: Mixture of Experts (MoE) — 30B total, only 3B active per token

Released: 2025-04

Similar models

qwen3.5-35b-a3bqwen3-8bqwen3-14b