Run unsloth/gemma-4-26B-A4B-it-GGUF locally

License: apache-2.0 ⬇ 1,465,954 ❤ 918
Parameters25.23B
Context262,144

unsloth/gemma-4-26B-A4B-it-GGUF is a large instruction-tuned chat model with 25.23 billion parameters, built on the gemma4 architecture. It is released under the apache-2.0 license and has been downloaded 1,465,954 times.

To run unsloth/gemma-4-26B-A4B-it-GGUF locally at a 4,096-token context, its quantized versions need between 5.13 GB (F16, lowest quality) and 52.17 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included.

For most users the best balance is F32, needing about 5.36 GB. That means unsloth/gemma-4-26B-A4B-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
F16 0.65 Very low 1.91 GB 2.42 GB 5.13 GB 209.7 t/s Fits in VRAM
F32 0.73 Very low 2.13 GB 2.42 GB 5.36 GB 187.5 t/s Fits in VRAM
IQ2_XXS 3.15 Low 9.24 GB 2.42 GB 12.46 GB 5.4 t/s Offload
IQ2_M 3.18 Low 9.33 GB 2.42 GB 12.55 GB 5.4 t/s Offload
Q2_K_XL 3.34 Fair 9.82 GB 2.42 GB 13.05 GB 5.1 t/s Offload
IQ3_S 3.58 Fair 10.51 GB 2.42 GB 13.74 GB 4.8 t/s Offload
IQ3_XXS 3.62 Fair 10.63 GB 2.42 GB 13.86 GB 4.7 t/s Offload
Q3_K_M 4.04 Fair 11.85 GB 2.42 GB 15.08 GB 4.2 t/s Offload
Q3_K_XL 4.09 Fair 12.02 GB 2.42 GB 15.24 GB 4.2 t/s Offload
IQ4_XS 4.31 Good 12.66 GB 2.42 GB 15.89 GB 3.9 t/s Offload
IQ4_NL 4.32 Good 12.68 GB 2.42 GB 15.9 GB 3.9 t/s Offload
Q4_K_S 5.23 Very good 15.36 GB 2.42 GB 18.58 GB 3.3 t/s Offload
Q4_K_M 5.37 Very good 15.78 GB 2.42 GB 19.01 GB 3.2 t/s Offload
Q4_K_XL 5.39 Very good 15.84 GB 2.42 GB 19.07 GB 3.2 t/s Offload
GGUF 5.39 Very good 15.84 GB 2.42 GB 19.07 GB 3.2 t/s Offload
Q5_K_S 5.98 Very good 17.56 GB 2.42 GB 20.78 GB 2.8 t/s Offload
Q5_K_M 6.71 Excellent 19.7 GB 2.42 GB 22.92 GB 2.5 t/s Offload
Q5_K_XL 6.73 Excellent 19.76 GB 2.42 GB 22.98 GB 2.5 t/s Offload
Q6_K 7.35 Excellent 21.58 GB 2.42 GB 24.8 GB Insufficient
Q6_K_XL 7.39 Excellent 21.7 GB 2.42 GB 24.92 GB Insufficient
Q8_0 8.66 Excellent 25.45 GB 2.42 GB 28.67 GB Insufficient
Q8_K_XL 8.76 Excellent 25.74 GB 2.42 GB 28.96 GB Insufficient
BF16 16.66 Excellent 48.95 GB 2.42 GB 52.17 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-4-26B-A4B-it-GGUF?

You need about 5.36 GB of VRAM to run unsloth/gemma-4-26B-A4B-it-GGUF entirely on the GPU using the F32 quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.

Can I run unsloth/gemma-4-26B-A4B-it-GGUF on an 8 GB GPU?

Yes. With 8 GB of VRAM you can run unsloth/gemma-4-26B-A4B-it-GGUF fully on the GPU using F32 (about 5.36 GB).

Can I run unsloth/gemma-4-26B-A4B-it-GGUF on a 16 GB GPU?

Yes. With 16 GB of VRAM you can run unsloth/gemma-4-26B-A4B-it-GGUF fully on the GPU using IQ4_NL (about 15.9 GB).

Can I run unsloth/gemma-4-26B-A4B-it-GGUF on a 24 GB GPU?

Yes. With 24 GB of VRAM you can run unsloth/gemma-4-26B-A4B-it-GGUF fully on the GPU using Q5_K_XL (about 22.98 GB).

What is the best quantization for unsloth/gemma-4-26B-A4B-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.