gemma-4-E2B-it - DASH-Q 2-bit GGUF
llama.cpp GGUF files of google/gemma-4-E2B-it quantized with DASH-Q, at several 2-bit-class sizes. Every file uses only standard llama.cpp tensor types (no tensor above 4 bits) and loads in any recent llama.cpp build.
| File | Type | Size | Bits / weight |
|---|---|---|---|
gemma-4-E2B-it-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 2.11 GB | 3.63 |
gemma-4-E2B-it-DASHQ-IQ2_M.gguf |
IQ2_M | 2.23 GB | 3.84 |
gemma-4-E2B-it-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 2.29 GB | 3.94 |
Text only: the vision tower is not included.
Quality vs. the unquantized model (KL divergence)
Raw-text perplexity is not a meaningful quality metric for this instruction/reasoning model (the unquantized model itself scores
poorly on WikiText-2 while generating normally), so quality is reported as KL divergence of the next-token distribution to the
unquantized reference (llama-perplexity --kl-divergence, WikiText-2, 4 x 2048 tokens):
| Type | Model | Size | Mean KLD | Median KLD | Top-1 token agreement |
|---|---|---|---|---|---|
| IQ2_XXS | DASH-Q IQ2_XXS | 2.11 GB | 3.03 | 1.95 | 44.9% |
| IQ2_M | unsloth UD-IQ2_M | 2.29 GB | 1.29 | 0.67 | 57.7% |
| IQ2_M | DASH-Q IQ2_M | 2.23 GB | 1.10 | 0.42 | 62.3% |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 2.97 GB | 1.28 | 0.66 | 58.5% |
| Q2_K_XL | unsloth UD-Q2_K_XL | 2.40 GB | 0.96 | 0.45 | 62.0% |
| Q2_K_XL | DASH-Q Q2_K_XL | 2.29 GB | 1.12 | 0.42 | 62.8% |
Usage
llama-cli -m gemma-4-E2B-it-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (google/gemma-4-E2B-it).