Qwen3 30B quants compared

27 GGUF builds by real file size, probed from bartowski/Qwen_Qwen3-30B-A3B-GGUF on Hugging Face (2026-09-02). 30B params, 3B active.

Download Qwen3 30B Q4_K_M (17.35 GB) — it fits 24 GB of memory with 16k context. With Ollama: ollama run qwen3:30b

Every Qwen3 30B quant by real file size

QuantWeights+ KV (16k)TotalFits comfortably inQuality
BF1656.9 GB4.0 GB60.9 GB96 GBFull precision (lossless)
Q8_030.25 GB4.0 GB34.3 GB48 GBNear-lossless
Q6_K_L23.52 GB4.0 GB27.5 GB32 GBExcellent
Q6_K23.38 GB4.0 GB27.4 GB32 GBExcellent
Q5_K_L20.43 GB4.0 GB24.4 GB32 GBVery high
Q5_K_M20.25 GB4.0 GB24.3 GB32 GBVery high
Q5_K_S19.65 GB4.0 GB23.6 GB32 GBVery high
Q4_K_M *17.35 GB4.0 GB21.4 GB24 GBHigh — the default pick
Q4_117.89 GB4.0 GB21.9 GB32 GBHigh
Q4_K_L17.57 GB4.0 GB21.6 GB24 GBHigh
Q4_K_S16.75 GB4.0 GB20.8 GB24 GBHigh
Q4_016.42 GB4.0 GB20.4 GB24 GBHigh
IQ4_NL16.19 GB4.0 GB20.2 GB24 GBHigh
IQ4_XS15.33 GB4.0 GB19.3 GB24 GBHigh
Q3_K_XL13.84 GB4.0 GB17.8 GB24 GBAcceptable — visible loss
Q3_K_L13.58 GB4.0 GB17.6 GB24 GBAcceptable — visible loss
IQ3_M13.11 GB4.0 GB17.1 GB24 GBAcceptable — visible loss
Q3_K_M13.11 GB4.0 GB17.1 GB24 GBAcceptable — visible loss
Q3_K_S12.51 GB4.0 GB16.5 GB24 GBAcceptable — visible loss
IQ3_XS11.86 GB4.0 GB15.9 GB24 GBAcceptable — visible loss
IQ3_XXS11.38 GB4.0 GB15.4 GB24 GBAcceptable — visible loss
Q2_K_L10.44 GB4.0 GB14.4 GB24 GBExperimental — not ranked — never recommended
Q2_K10.16 GB4.0 GB14.2 GB16 GBExperimental — not ranked — never recommended
IQ2_M9.71 GB4.0 GB13.7 GB16 GBExperimental — not ranked — never recommended
IQ2_S8.59 GB4.0 GB12.6 GB16 GBExperimental — not ranked — never recommended
IQ2_XS8.51 GB4.0 GB12.5 GB16 GBExperimental — not ranked — never recommended
IQ2_XXS7.59 GB4.0 GB11.6 GB16 GBExperimental — not ranked — never recommended

* default pick. Weights = real GGUF file sizes from bartowski/Qwen_Qwen3-30B-A3B-GGUF (probed 2026-09-02). KV = fp16 estimate; a q8_0 cache roughly halves it. "Comfortable" = weights + KV within 90% of memory.

Best Qwen3 30B quant by memory

MemoryRecommended quantTotal (16k ctx)
24 GBQ4_K_M21.4 GB
32 GBQ6_K27.4 GB
48 GBQ8_034.3 GB
96 GBBF1660.9 GB

Why we don't rank Qwen3 30B's 2-bit quants

Quants at 2 bits per weight or below (Q2_K, IQ2, IQ1, TQ1) cut file size by roughly half versus Q4, but the quality collapse is steep and non-linear: perplexity spikes, instruction-following degrades, and hallucinations rise. A model that answers faster but wrong is not a smaller model — it is a worse one. ModelFit lists these builds for completeness but never ranks or recommends them.

Frequently asked questions

What is the best quantization of Qwen3 30B?

Q4_K_M is the default pick: 17.35 GB of weights, high — the default pick quality, fitting comfortably in 24 GB of memory (weights + 16k context KV-cache). Go Q6_K or Q8_0 if you have headroom.

How much memory does Qwen3 30B need?

At Q4_K_M, Qwen3 30B needs 17.35 GB for the weights plus ~4.0 GB of KV-cache at 16k context — about 21.4 GB total, so a 24 GB card or Mac (90% usable budget) runs it comfortably.

Should I use a Q2_K or IQ2 quant of Qwen3 30B?

No. Qwen3 30B at 2 bits per weight is a visibly worse model — quality collapse at that bitrate is steep, not gradual. If only a 2-bit build fits your memory, run a smaller model at Q4_K_M instead. ModelFit lists these builds but never recommends them.

Cite this page

ModelFit: Qwen3 30B quantization comparison (real GGUF file sizes).
https://modelfit.io/quant-compare/qwen3-30b-a3b/ (data probed 2026-09-02, CC BY 4.0).