Qwen3-Coder-30B-A3B-Instruct GGUF size and VRAM requirements

License: apache-2.0 ⬇ 7,961,172 ❤ 894
Parameters30.53B
Context262,144

unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF is a very large code-focused language model with 30.53 billion parameters, built on the qwen3moe architecture. It is released under the apache-2.0 license and has been downloaded 7,961,172 times.

Qwen3-Coder-30B-A3B-Instruct-GGUF is a Mixture-of-Experts model with 128 experts, of which 8 are active on each token. Routing only 8 of 128 experts makes it noticeably faster than a dense model of the same size — but every expert still has to be held in memory, so the numbers below are set by the full parameter count, not the active one. It is built from 48 transformer layers, a hidden size of 2,048, 32 attention heads. It uses grouped-query attention (32 query heads sharing 4 key/value heads), which already trims KV-cache memory compared with full multi-head attention.

To run unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF locally at a 4,096-token context, its quantized versions need between 8.63 GB (TQ1_0, lowest quality) and 58.07 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included. Context length drives that memory directly: at 4,096 tokens the KV cache for Qwen3-Coder-30B-A3B-Instruct-GGUF is about 0.38 GB, rising to roughly 24.0 GB at its full 262,144-token context. Shorter prompts free up memory for a higher-quality quantization.

For most users the best balance is Q5_K_XL, needing about 21.42 GB. That means unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF fits entirely in the VRAM of a 10 GB GPU or larger, running fully on the GPU.

Available GGUF quantizations for unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF include TQ1_0, IQ1_S, IQ1_M, IQ2_XXS, IQ2_M, Q2_K, Q2_K_L, Q2_K_XL, IQ3_XXS, Q3_K_S, Q3_K_XL, Q3_K_M, IQ4_XS, IQ4_NL, Q4_0, Q4_K_S, Q4_K_XL, Q4_K_M, Q4_1, Q5_K_S, Q5_K_M, Q5_K_XL, Q6_K, Q6_K_XL, Q8_0, Q8_K_XL, BF16. The model supports a native context length of up to 262,144 tokens; a longer context grows the KV cache and the memory needed.

→ Guide: How much VRAM do you need?

GGUF file size and memory by quantization

Compare real GGUF weight sizes, estimated KV cache and total memory for Q4, Q5, Q8 and every quantization published in this repository.

Quant.Bits QualityWeights KVTotal Speed~Verdict
TQ1_0 2.1 Very low 7.46 GB 0.38 GB 8.63 GB 6.7 t/s Offload
IQ1_S 2.34 Very low 8.3 GB 0.38 GB 9.48 GB 6.0 t/s Offload
IQ1_M 2.52 Very low 8.97 GB 0.38 GB 10.14 GB 5.6 t/s Offload
IQ2_XXS 2.71 Low 9.62 GB 0.38 GB 10.8 GB 5.2 t/s Offload
IQ2_M 2.84 Low 10.09 GB 0.38 GB 11.27 GB 5.0 t/s Offload
Q2_K 2.95 Low 10.49 GB 0.38 GB 11.66 GB 4.8 t/s Offload
Q2_K_L 2.97 Low 10.55 GB 0.38 GB 11.73 GB 4.7 t/s Offload
Q2_K_XL 3.09 Low 10.98 GB 0.38 GB 12.15 GB 4.6 t/s Offload
IQ3_XXS 3.37 Fair 11.97 GB 0.38 GB 13.14 GB 4.2 t/s Offload
Q3_K_S 3.48 Fair 12.38 GB 0.38 GB 13.55 GB 4.0 t/s Offload
Q3_K_XL 3.62 Fair 12.86 GB 0.38 GB 14.03 GB 3.9 t/s Offload
Q3_K_M 3.85 Fair 13.7 GB 0.38 GB 14.88 GB 3.6 t/s Offload
IQ4_XS 4.29 Good 15.25 GB 0.38 GB 16.43 GB 3.3 t/s Offload
IQ4_NL 4.54 Good 16.12 GB 0.38 GB 17.3 GB 3.1 t/s Offload
Q4_0 4.55 Good 16.19 GB 0.38 GB 17.36 GB 3.1 t/s Offload
Q4_K_S 4.57 Good 16.26 GB 0.38 GB 17.43 GB 3.1 t/s Offload
Q4_K_XL 4.63 Good 16.45 GB 0.38 GB 17.63 GB 3.0 t/s Offload
Q4_K_M 4.86 Good 17.28 GB 0.38 GB 18.46 GB 2.9 t/s Offload
Q4_1 5.03 Very good 17.87 GB 0.38 GB 19.05 GB 2.8 t/s Offload
Q5_K_S 5.52 Very good 19.63 GB 0.38 GB 20.81 GB 2.5 t/s Offload
Q5_K_M 5.69 Very good 20.23 GB 0.38 GB 21.41 GB 2.5 t/s Offload
Q5_K_XL 5.7 Very good 20.25 GB 0.38 GB 21.42 GB 2.5 t/s Offload
Q6_K 6.57 Excellent 23.37 GB 0.38 GB 24.54 GB Insufficient
Q6_K_XL 6.9 Excellent 24.53 GB 0.38 GB 25.71 GB Insufficient
Q8_0 8.51 Excellent 30.25 GB 0.38 GB 31.43 GB Insufficient
Q8_K_XL 9.43 Excellent 33.52 GB 0.38 GB 34.69 GB Insufficient
BF16 16.01 Excellent 56.9 GB 0.38 GB 58.07 GB Insufficient

KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.

Frequently asked questions

What kind of model is unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF?

unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF is a code-focused language model with 30.53 billion parameters, based on the qwen3moe architecture. It is released under the apache-2.0 license and distributed as GGUF files for local inference.

Is unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF a Mixture-of-Experts (MoE) model?

Yes. unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF is a Mixture-of-Experts model with 128 experts, of which 8 are activated per token. That makes it faster than a dense model of the same size, but all 128 experts must be loaded into memory, so the VRAM/RAM it needs is driven by the total parameter count, not the active one.

How much VRAM do you need to run unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF?

You need about 9.48 GB of VRAM to run unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF entirely on the GPU using the IQ1_S quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.

Can I run unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF on an 8 GB GPU?

Partially. unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF only fits on an 8 GB GPU by offloading part of it to system RAM (with Q5_K_XL), which runs but is slower.

Can I run unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF on a 16 GB GPU?

Yes. With 16 GB of VRAM you can run unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF fully on the GPU using Q3_K_M (about 14.88 GB).

Can I run unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF on a 24 GB GPU?

Yes. With 24 GB of VRAM you can run unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF fully on the GPU using Q5_K_XL (about 21.42 GB).

What context length does unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF support?

unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF supports a native context length of up to 262,144 tokens. A longer context grows the KV cache, so it increases the memory needed to run the model.

What is the best quantization for unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF?

For unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF, a strong default is Q4_K_M, which needs about 18.46 GB and keeps most of the quality while roughly halving the memory versus 8-bit. With VRAM to spare, Q5_K_M or Q6_K add a little more quality; if you are tight on memory, a smaller quantization still runs. Pick the highest quantization that fits your VRAM.