Run unsloth/gemma-4-E2B-it-GGUF locally
unsloth/gemma-4-E2B-it-GGUF is a mid-size instruction-tuned chat model with 4.65 billion parameters, built on the gemma4 architecture. It is released under the apache-2.0 license and has been downloaded 696,110 times.
To run unsloth/gemma-4-E2B-it-GGUF locally at a 4,096-token context, its quantized versions need between 0.9 GB (GGUF, lowest quality) and 10.55 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included.
For most users the best balance is Q8_K_XL, needing about 5.72 GB. That means unsloth/gemma-4-E2B-it-GGUF fits entirely in the VRAM of a 6 GB GPU or larger, running fully on the GPU.
All quantizations
| Quant. | Bits | Quality | Weights | KV | Total | Speed~ | Verdict |
|---|---|---|---|---|---|---|---|
| GGUF | 0.17 | Very low | 0.09 GB | 0.0 GB | 0.9 GB | 4390.8 t/s | Fits in VRAM |
| F16 | 1.99 | Very low | 1.08 GB | 0.0 GB | 1.88 GB | 371.6 t/s | Fits in VRAM |
| F32 | 3.28 | Low | 1.77 GB | 0.0 GB | 2.58 GB | 225.7 t/s | Fits in VRAM |
| IQ2_M | 3.94 | Fair | 2.13 GB | 0.0 GB | 2.94 GB | 187.5 t/s | Fits in VRAM |
| IQ3_XXS | 4.08 | Fair | 2.21 GB | 0.0 GB | 3.01 GB | 181.0 t/s | Fits in VRAM |
| Q2_K_XL | 4.14 | Fair | 2.24 GB | 0.0 GB | 3.04 GB | 178.7 t/s | Fits in VRAM |
| Q3_K_S | 4.21 | Good | 2.28 GB | 0.0 GB | 3.08 GB | 175.6 t/s | Fits in VRAM |
| Q3_K_M | 4.37 | Good | 2.36 GB | 0.0 GB | 3.17 GB | 169.3 t/s | Fits in VRAM |
| Q3_K_XL | 5.03 | Very good | 2.72 GB | 0.0 GB | 3.53 GB | 146.9 t/s | Fits in VRAM |
| IQ4_XS | 5.14 | Very good | 2.78 GB | 0.0 GB | 3.58 GB | 143.9 t/s | Fits in VRAM |
| IQ4_NL | 5.23 | Very good | 2.83 GB | 0.0 GB | 3.64 GB | 141.2 t/s | Fits in VRAM |
| Q4_0 | 5.24 | Very good | 2.83 GB | 0.0 GB | 3.64 GB | 141.2 t/s | Fits in VRAM |
| Q4_K_S | 5.24 | Very good | 2.83 GB | 0.0 GB | 3.64 GB | 141.1 t/s | Fits in VRAM |
| Q4_K_M | 5.35 | Very good | 2.89 GB | 0.0 GB | 3.7 GB | 138.2 t/s | Fits in VRAM |
| Q4_1 | 5.43 | Very good | 2.94 GB | 0.0 GB | 3.74 GB | 136.1 t/s | Fits in VRAM |
| Q4_K_XL | 5.48 | Very good | 2.97 GB | 0.0 GB | 3.77 GB | 134.9 t/s | Fits in VRAM |
| Q5_K_S | 5.72 | Very good | 3.09 GB | 0.0 GB | 3.9 GB | 129.3 t/s | Fits in VRAM |
| Q5_K_M | 5.78 | Very good | 3.13 GB | 0.0 GB | 3.93 GB | 128.0 t/s | Fits in VRAM |
| Q5_K_XL | 7.39 | Excellent | 4.0 GB | 0.0 GB | 4.8 GB | 100.0 t/s | Fits in VRAM |
| Q6_K | 7.75 | Excellent | 4.19 GB | 0.0 GB | 5.0 GB | 95.4 t/s | Fits in VRAM |
| Q6_K_XL | 8.11 | Excellent | 4.39 GB | 0.0 GB | 5.19 GB | 91.2 t/s | Fits in VRAM |
| Q8_0 | 8.86 | Excellent | 4.79 GB | 0.0 GB | 5.6 GB | 83.5 t/s | Fits in VRAM |
| Q8_K_XL | 9.09 | Excellent | 4.92 GB | 0.0 GB | 5.72 GB | 81.3 t/s | Fits in VRAM |
| BF16 | 18.02 | Excellent | 9.75 GB | 0.0 GB | 10.55 GB | 5.1 t/s | Offload |
KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.
Frequently asked questions
How much VRAM do you need to run unsloth/gemma-4-E2B-it-GGUF?
You need about 5.72 GB of VRAM to run unsloth/gemma-4-E2B-it-GGUF entirely on the GPU using the Q8_K_XL quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.
Can I run unsloth/gemma-4-E2B-it-GGUF on an 8 GB GPU?
Yes. With 8 GB of VRAM you can run unsloth/gemma-4-E2B-it-GGUF fully on the GPU using Q8_K_XL (about 5.72 GB).
Can I run unsloth/gemma-4-E2B-it-GGUF on a 16 GB GPU?
Yes. With 16 GB of VRAM you can run unsloth/gemma-4-E2B-it-GGUF fully on the GPU using BF16 (about 10.55 GB).
Can I run unsloth/gemma-4-E2B-it-GGUF on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run unsloth/gemma-4-E2B-it-GGUF fully on the GPU using BF16 (about 10.55 GB).
What is the best quantization for unsloth/gemma-4-E2B-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.