Run unsloth/gemma-4-12b-it-GGUF locally

License: apache-2.0 ⬇ 1,390,513 ❤ 710
Parameters11.91B
Context262,144

unsloth/gemma-4-12b-it-GGUF is a large instruction-tuned chat model with 11.91 billion parameters, built on the gemma4 architecture. It is released under the apache-2.0 license and has been downloaded 1,390,513 times.

To run unsloth/gemma-4-12b-it-GGUF locally at a 4,096-token context, its quantized versions need between 2.66 GB (F32, lowest quality) and 25.63 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included.

For most users the best balance is Q3_K_M, needing about 7.77 GB. That means unsloth/gemma-4-12b-it-GGUF fits entirely in the VRAM of a 6 GB GPU or larger, running fully on the GPU.

→ Guide: How much VRAM do you need?

All quantizations

Quant.Bits QualityWeights KVTotal Speed~Verdict
F32 0.14 Very low 0.2 GB 1.67 GB 2.66 GB 2049.9 t/s Fits in VRAM
GGUF 0.31 Very low 0.43 GB 1.67 GB 2.9 GB 923.4 t/s Fits in VRAM
F16 0.7 Very low 0.97 GB 1.67 GB 3.43 GB 414.3 t/s Fits in VRAM
IQ2_M 2.83 Low 3.92 GB 1.67 GB 6.39 GB 101.9 t/s Fits in VRAM
IQ3_XXS 3.12 Low 4.32 GB 1.67 GB 6.79 GB 92.6 t/s Fits in VRAM
Q2_K_XL 3.13 Low 4.34 GB 1.67 GB 6.81 GB 92.1 t/s Fits in VRAM
Q3_K_S 3.45 Fair 4.78 GB 1.67 GB 7.25 GB 83.6 t/s Fits in VRAM
Q3_K_M 3.83 Fair 5.3 GB 1.67 GB 7.77 GB 75.4 t/s Fits in VRAM
Q3_K_XL 4.05 Fair 5.61 GB 1.67 GB 8.07 GB 8.9 t/s Offload
IQ4_XS 4.28 Good 5.94 GB 1.67 GB 8.4 GB 8.4 t/s Offload
IQ4_NL 4.51 Good 6.26 GB 1.67 GB 8.72 GB 8.0 t/s Offload
Q4_0 4.53 Good 6.28 GB 1.67 GB 8.74 GB 8.0 t/s Offload
Q4_K_S 4.54 Good 6.3 GB 1.67 GB 8.77 GB 7.9 t/s Offload
Q4_K_M 4.78 Good 6.63 GB 1.67 GB 9.1 GB 7.5 t/s Offload
Q4_K_XL 4.95 Good 6.86 GB 1.67 GB 9.33 GB 7.3 t/s Offload
Q4_1 4.97 Good 6.89 GB 1.67 GB 9.35 GB 7.3 t/s Offload
Q5_K_S 5.51 Very good 7.64 GB 1.67 GB 10.11 GB 6.5 t/s Offload
Q5_K_M 5.65 Very good 7.84 GB 1.67 GB 10.3 GB 6.4 t/s Offload
Q5_K_XL 5.78 Very good 8.01 GB 1.67 GB 10.48 GB 6.2 t/s Offload
Q6_K 6.57 Excellent 9.11 GB 1.67 GB 11.58 GB 5.5 t/s Offload
Q6_K_XL 7.18 Excellent 9.95 GB 1.67 GB 12.42 GB 5.0 t/s Offload
Q8_0 8.82 Excellent 12.23 GB 1.67 GB 14.7 GB 4.1 t/s Offload
Q8_K_XL 9.16 Excellent 12.7 GB 1.67 GB 15.17 GB 3.9 t/s Offload
BF16 16.71 Excellent 23.16 GB 1.67 GB 25.63 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 unsloth/gemma-4-12b-it-GGUF?

You need about 3.43 GB of VRAM to run unsloth/gemma-4-12b-it-GGUF entirely on the GPU using the F16 quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.

Can I run unsloth/gemma-4-12b-it-GGUF on an 8 GB GPU?

Yes. With 8 GB of VRAM you can run unsloth/gemma-4-12b-it-GGUF fully on the GPU using Q3_K_M (about 7.77 GB).

Can I run unsloth/gemma-4-12b-it-GGUF on a 16 GB GPU?

Yes. With 16 GB of VRAM you can run unsloth/gemma-4-12b-it-GGUF fully on the GPU using Q8_K_XL (about 15.17 GB).

Can I run unsloth/gemma-4-12b-it-GGUF on a 24 GB GPU?

Yes. With 24 GB of VRAM you can run unsloth/gemma-4-12b-it-GGUF fully on the GPU using Q8_K_XL (about 15.17 GB).

What is the best quantization for unsloth/gemma-4-12b-it-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.