CoolFace
Modelpublic

AtomicChat/gemma-4-E4B-it-GGUF

sourceHugging Faceapache-2.0updated 2mo agoView on Hugging Face
1likes3.8kdownloads
Model Card

<center>

<div style="display:flex; justify-content:center; align-items:center; gap:2%; max-width:560px; margin:0 auto;"> <a href="https://atomic.chat" style="flex:0 1 auto; min-width:0;"><img src="https://huggingface.co/AtomicChat/gemma-4-E4B-it-GGUF/resolve/main/pillatomicv3.png" alt="Atomic Chat" style="width:100%; height:auto; max-width:186px;"></a> <a href="https://discord.gg/8wGSsvmg4V" style="flex:0 1 auto; min-width:0;"><img src="https://huggingface.co/AtomicChat/gemma-4-E4B-it-GGUF/resolve/main/pilldiscordv3.png" alt="Join Discord" style="width:100%; height:auto; max-width:184px;"></a> <a href="https://github.com/AtomicBot-ai/Atomic-Chat" style="flex:0 1 auto; min-width:0;"><img src="https://huggingface.co/AtomicChat/gemma-4-E4B-it-GGUF/resolve/main/pillgithubv3.png" alt="GitHub" style="width:100%; height:auto; max-width:141px;"></a> </div>

<br/>

<img src="https://huggingface.co/AtomicChat/gemma-4-E4B-it-GGUF/resolve/main/hero.png" alt="Gemma 4 E4B" style="width:100%; max-width:100%; height:auto; margin-bottom:0.6em;"/>

<div style="display:flex; justify-content:center; gap:0.5em;"> <a href="https://huggingface.co/google/gemma-4-E4B-it"><strong>Base model: google/gemma-4-E4B-it</strong></a> </div> </center>

Gemma 4 E4B, self-quantized to GGUF by Atomic Chat. Built straight from Google's original weights with a per-tensor importance matrix, so this is not a repack of somebody else's files. Runs fully offline.

Highlights

  • —4.5B effective (8B with embeddings) parameters: the weights this repo quantizes.
  • —Context length: 128K tokens, as published by Google.
  • —42 layers: Dense decoder, hybrid sliding-window (512) and global attention.
  • —Modalities: Text, Image, Audio.
  • —Full imatrix ladder: every quant is calibrated with an importance matrix, published here alongside the quants.
  • —Reasoning: All models in the family are designed as highly capable reasoners, with configurable thinking modes.
  • —Diverse & Efficient Architectures: Offers Dense and Mixture-of-Experts (MoE) variants of different sizes for scalable deployment.
[!NOTE] These GGUFs are self-quantized from the original weights, not a repack. The importance matrix keeps low-bit quants closer to the full-precision model.
[!IMPORTANT] Always pass --jinja so the Gemma 4 E4B chat template is applied. Without it the model can emit malformed turns.

Model Overview

PropertyValue
Base modelgoogle/gemma-4-E4B-it
Parameters4.5B effective (8B with embeddings)
Layers42
Sliding window512 tokens
Context length128K tokens
Vocabulary262K
ModalitiesText, Image, Audio
ArchitectureDense decoder, hybrid sliding-window (512) and global attention, 8 attention heads over 2 KV heads, Gemma4ForConditionalGeneration
This repoGGUF quants (imatrix) and a vision mmproj; the importance matrix is published here as imatrix-coding.gguf. Quants: Q2_K, IQ3_M, Q3_K_M, Q3_K_L, IQ4_XS, Q4_K_S, Q4_K_M, Q5_K_S, Q5_K_M, UD-Q4_K_XL, Q6_K, Q8_0
[!NOTE] Gemma 4 E4B is multimodal. This repo ships the `mmproj-gemma4-e4b-it-f16.gguf` vision projector. With -hf it is pulled automatically; otherwise pass --mmproj. Use llama-mtmd-cli or llama-server to feed images.

<img src="https://huggingface.co/AtomicChat/gemma-4-E4B-it-GGUF/resolve/main/benchmark.png" alt="Gemma 4 E4B benchmark scores" style="width:100%; max-width:900px;"/>

Scores are Google's published results for the base google/gemma-4-E4B-it, not our own measurements. Quantization preserves the large majority of this; Q4_K_M and up stay close to full precision.

Choosing a quant

QuantSizeNotes
Q2_K4.4 GBSmallest K-quant. Minimal RAM, clear quality drop.
IQ3_M4.7 GBBeats Q3 at a similar size thanks to imatrix. Best low-RAM pick.
Q3_K_M4.9 GBLow quality but usable.
Q3_K_L5.0 GBA step above Q3KM.
IQ4_XS5.1 GBExcellent quality for size. Recommended low-bit.
Q4_K_S5.2 GBCompact 4-bit, fast.
`Q4_K_M`5.3 GBRecommended default. Best balance of size, speed and quality.
Q5_K_S5.7 GBHigher quality, slightly more compact than Q5KM.
Q5_K_M5.8 GBHigher quality, low loss.
UD-Q4_K_XL6.2 GBDynamic. Embeddings and output kept at Q8_0 for higher quality at a Q4 footprint.
Q6_K6.2 GBNear lossless, noticeably lighter than Q8_0.
Q8_08.0 GBEffectively lossless, reference quality.
[!TIP] Pick the largest file that fits your (V)RAM with room for context. Q4_K_M or UD-Q4_K_XL is the sweet spot for most setups; Q6_K or Q8_0 for maximum fidelity.

Get started

Run Gemma 4 E4B locally with:

  • —[Atomic Chat](https://atomic.chat): the easiest path. Open the app, search AtomicChat/gemma-4-E4B-it-GGUF, pick a quant, hit Use this model.
  • —llama.cpp: llama-server -hf AtomicChat/gemma-4-E4B-it-GGUF:Q4_K_M --jinja -c 8192
  • —Ollama: ollama run hf.co/AtomicChat/gemma-4-E4B-it-GGUF:Q4_K_M
  • —LM Studio / Jan: search the repo id, download any quant.

Best practices

ParameterValue
temperature1.0
top_p0.95
top_k64

Google's recommended sampling configuration for google/gemma-4-E4B-it. Pass images through llama-mtmd-cli or llama-server with the projector.

Run in llama.cpp

bash
git clone https://github.com/ggml-org/llama.cpp
cmake llama.cpp -B llama.cpp/build -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=ON
cmake --build llama.cpp/build --config Release -j --target llama-cli llama-server
bash
./llama.cpp/build/bin/llama-server \
    -hf AtomicChat/gemma-4-E4B-it-GGUF:Q4_K_M \
    --jinja -ngl 99 -c 8192 -fa on

How these were made

  1. 1.Download google/gemma-4-E4B-it (original weights).
  2. 2.Convert to f16 GGUF with llama.cpp.
  3. 3.Build an importance matrix over our calibration corpus, published here as imatrix-coding.gguf.
  4. 4.Quantize the ladder with --imatrix.
  5. 5.UD-Q4_K_XL additionally pins the token-embedding and output tensors to Q8_0.

License

Original model by Google, released under the Apache 2.0 license. Full terms: Apache 2.0. Quantized by Atomic Chat.