LocalForge
Voltar ao guia
MODEL / ALIBABA QWEN

Qwen3 30B-A3B

All 30B parameters must be stored even though roughly 3B are active per token.

Fonte primária
Parâmetros totais30B3B active / token
Memória estimada~22.5 GBGGUF Q4 MoE
Contexto32K+Text
LicenseApache-2.0100+ languages and dialects
01 / IDEAL PARA
  • reasoning
  • coding
  • efficient MoE
02 / AJUSTE ESTIMADO
Apple silicon · 32 GBTightNVIDIA · 24 GBTightApple silicon · 64 GBComfortableWorkstation · 48 GBComfortableDatacenter · 80 GBComfortable
03 / RODAR COM
OllamaEasyollama run qwen3:30b-a3b
llama.cppMediumllama-server -m /path/to/verified-model.gguf --port 8080
LM StudioEasySearch “Qwen/Qwen3-30B-A3B” → Load a Q4/INT4 build → Start Server
MLX LMMediummlx_lm.chat --model Qwen/Qwen3-30B-A3B
Atenção

All 30B parameters must be stored even though roughly 3B are active per token.