Skip to main content
Local LLM Guide

Model

Qwen3 30B-A3B

Start with a supported compressed build and short context. 29.3GB RAM estimate. Best use: Strong local reasoning when you want more capability than dense 14B without running a full 32B dense model at every token.

30.5B total / 3.3B activeReasoning

Best use

Strong local reasoning when you want more capability than dense 14B without running a full 32B dense model at every token. · Math, coding, and logical reasoning workloads that benefit from Qwen3 thinking mode.

Memory planning

29.3GB RAM estimate

Expected pace

Depends on your hardware, app, and conversation length

License

Apache-2.0; read the model’s terms before use

Who should use this model?

  • Strong local reasoning when you want more capability than dense 14B without running a full 32B dense model at every token.
  • Math, coding, and logical reasoning workloads that benefit from Qwen3 thinking mode.
  • Local agent experiments where Qwen3 tool/agent behavior is useful and memory can handle the full MoE checkpoint.

Choose something else when

  • You need an 8GB or 16GB RAM laptop target.
  • You need the smallest possible download; MoE still stores all experts on disk.
  • Your runtime lacks stable Qwen3-MoE support.

What hardware makes sense?

The download is only part of the memory needed. Leave room for the app, the input it processes, and other open programs. For chat models, context includes the instructions, conversation, and answer. Longer context needs more memory.

CPU only
Possible but slow
8GB system memory
Use a smaller model for this memory limit.
16GB system memory
Use a smaller model for this memory limit.
32GB system memory
A possible starting point with a compressed model and short conversation; check the exact file size.
8GB GPU memory
Use a smaller model for this memory limit.
12GB GPU memory
Use a smaller model for this memory limit.
24GB GPU memory
Memory or response time may limit use. Try a smaller model first.
Apple unified memory
Good 32GB plus

Model facts that change the decision

Parameters
30.5B total / 3.3B active
Input
Text
Published context limit
32,768 tokens
Model format
Safetensors · GGUF · Ollama
Suggested compression
Q4_K_M for practical local use; Q5_K_M if 32GB+ unified memory or ample VRAM is available.
Typical Q4 file
18.6GB before runtime and context overhead
Languages
Chinese · English · French · Spanish · Portuguese · German · Italian · Russian

First setup

Ollama

ollama run qwen3:30b

LM Studio: Search for `Qwen/Qwen3-30B-A3B-GGUF` in LM Studio and select Q4_K_M or Q5_K_M.

llama.cpp

Use a recent llama.cpp build with Qwen3-MoE support, for example llama-server -hf Qwen/Qwen3-30B-A3B-GGUF:Q4_K_M; control thinking mode with /think and /no_think.

Conditions that change the answer

  • RAM and VRAM figures are estimates from GGUF file size plus conservative runtime and conversation memory overhead; exact memory depends on context length, batch size, backend, and GPU offload.
  • For MoE models, the active parameter count reflects the official activated-parameter count per token, while download and memory needs reflect the full stored parameter set.
  • Qwen3 model cards state 32,768 native context and 131,072 tokens with YaRN. Use the native limit unless you configure the longer setting.
  • Ollama currently presents newer Qwen3 30B packaging as `qwen3:30b`; the official Hugging Face artifact for this requested target is `Qwen3-30B-A3B`.

Read the controlling license text →