copywriter-gemma4-31b-i1 GGUF size and VRAM requirements
mradermacher/copywriter-gemma4-31b-i1-GGUF is a very large language model with 30.7 billion parameters, built on the gemma4 architecture. It is released under the gemma license and has been downloaded 12,467 times.
To run mradermacher/copywriter-gemma4-31b-i1-GGUF locally at a 4,096-token context, its quantized versions need between 3.49 GB (GGUF, lowest quality) and 26.94 GB (Q6_K, highest quality) of memory, weights plus KV cache and a system margin included.
For most users the best balance is GGUF, needing about 3.49 GB. That means mradermacher/copywriter-gemma4-31b-i1-GGUF fits entirely in the VRAM of a 6 GB GPU or larger, running fully on the GPU.
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 | Quality | Weights | KV | Total | Speed~ | Verdict |
|---|---|---|---|---|---|---|---|
| GGUF | — | Very low | 0.01 GB | 2.67 GB | 3.49 GB | 31227.2 t/s | Fits in VRAM |
| IQ1_S | 1.87 | Very low | 6.67 GB | 2.67 GB | 10.14 GB | 7.5 t/s | Offload |
| IQ1_M | 2.01 | Very low | 7.2 GB | 2.67 GB | 10.67 GB | 6.9 t/s | Offload |
| IQ2_XXS | 2.26 | Very low | 8.08 GB | 2.67 GB | 11.55 GB | 6.2 t/s | Offload |
| IQ2_XS | 2.48 | Very low | 8.88 GB | 2.67 GB | 12.35 GB | 5.6 t/s | Offload |
| IQ2_S | 2.65 | Low | 9.46 GB | 2.67 GB | 12.93 GB | 5.3 t/s | Offload |
| IQ2_M | 2.85 | Low | 10.17 GB | 2.67 GB | 13.64 GB | 4.9 t/s | Offload |
| Q2_K_S | 2.86 | Low | 10.22 GB | 2.67 GB | 13.7 GB | 4.9 t/s | Offload |
| Q2_K | 3.11 | Low | 11.1 GB | 2.67 GB | 14.57 GB | 4.5 t/s | Offload |
| IQ3_XXS | 3.15 | Low | 11.25 GB | 2.67 GB | 14.72 GB | 4.4 t/s | Offload |
| IQ3_XS | 3.41 | Fair | 12.17 GB | 2.67 GB | 15.65 GB | 4.1 t/s | Offload |
| IQ3_S | 3.59 | Fair | 12.82 GB | 2.67 GB | 16.29 GB | 3.9 t/s | Offload |
| Q3_K_S | 3.59 | Fair | 12.82 GB | 2.67 GB | 16.29 GB | 3.9 t/s | Offload |
| IQ3_M | 3.76 | Fair | 13.43 GB | 2.67 GB | 16.91 GB | 3.7 t/s | Offload |
| Q3_K_M | 3.98 | Fair | 14.24 GB | 2.67 GB | 17.71 GB | 3.5 t/s | Offload |
| Q3_K_L | 4.33 | Good | 15.49 GB | 2.67 GB | 18.96 GB | 3.2 t/s | Offload |
| IQ4_XS | 4.36 | Good | 15.59 GB | 2.67 GB | 19.06 GB | 3.2 t/s | Offload |
| Q4_0 | 4.61 | Good | 16.49 GB | 2.67 GB | 19.96 GB | 3.0 t/s | Offload |
| Q4_K_S | 4.63 | Good | 16.54 GB | 2.67 GB | 20.02 GB | 3.0 t/s | Offload |
| Q4_K_M | 4.87 | Good | 17.4 GB | 2.67 GB | 20.88 GB | 2.9 t/s | Offload |
| Q4_1 | 5.08 | Very good | 18.14 GB | 2.67 GB | 21.62 GB | 2.8 t/s | Offload |
| Q5_K_S | 5.55 | Very good | 19.85 GB | 2.67 GB | 23.32 GB | 2.5 t/s | Offload |
| Q5_K_M | 5.69 | Very good | 20.35 GB | 2.67 GB | 23.82 GB | 2.5 t/s | Offload |
| Q6_K | 6.57 | Excellent | 23.47 GB | 2.67 GB | 26.94 GB | — | Insufficient |
KV cache estimated (architecture unavailable). Speed is a rough estimate bounded by memory bandwidth.
Frequently asked questions
How much VRAM do you need to run mradermacher/copywriter-gemma4-31b-i1-GGUF?
You need about 3.49 GB of VRAM to run mradermacher/copywriter-gemma4-31b-i1-GGUF entirely on the GPU using the GGUF quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.
Can I run mradermacher/copywriter-gemma4-31b-i1-GGUF on an 8 GB GPU?
Yes. With 8 GB of VRAM you can run mradermacher/copywriter-gemma4-31b-i1-GGUF fully on the GPU using GGUF (about 3.49 GB).
Can I run mradermacher/copywriter-gemma4-31b-i1-GGUF on a 16 GB GPU?
Yes. With 16 GB of VRAM you can run mradermacher/copywriter-gemma4-31b-i1-GGUF fully on the GPU using IQ3_XS (about 15.65 GB).
Can I run mradermacher/copywriter-gemma4-31b-i1-GGUF on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run mradermacher/copywriter-gemma4-31b-i1-GGUF fully on the GPU using Q5_K_M (about 23.82 GB).
What is the best quantization for mradermacher/copywriter-gemma4-31b-i1-GGUF?
If memory allows, higher bits-per-weight means better quality. A common sweet spot is a Q4_K_M or Q5_K_M quantization, which keeps most of the quality while roughly halving the memory versus 8-bit. Pick the highest quantization that still fits in your VRAM.