AtomicChat/gemma-4-E2B-it-assistant-GGUF

Base model: google/gemma-4-E2B-it-assistant

Gemma 4 E2B It Assistant, 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

• 2.3B effective (5.1B with embeddings) parameters: the weights this repo quantizes.
• Context length: 128K tokens, as published by Google.
• 35 layers: Dense decoder, hybrid sliding-window (512) and global attention.
• Modalities: the base model handles Text, Image, Audio; this repo ships text-only quants, it carries no vision projector.
• Full imatrix ladder: every quant is calibrated with an importance matrix.
• 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.

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.

Always pass --jinja so the Gemma 4 E2B It Assistant chat template is applied. Without it the model can emit malformed turns.

Model Overview

Property	Value
Base model	google/gemma-4-E2B-it-assistant
Parameters	2.3B effective (5.1B with embeddings)
Layers	35
Sliding window	512 tokens
Context length	128K tokens
Vocabulary	262K
Modalities	Text, Image, Audio in the base model; text only in this repo, it ships no vision projector
Architecture	Dense decoder, hybrid sliding-window (512) and global attention, 4 attention heads over 1 KV head, Gemma4AssistantForCausalLM
This repo	GGUF quants (imatrix). Quants: Q4_K_S, Q4_K_M, Q5_K_M, Q8_0, F16

Benchmarks

Benchmark	Score
MMLU Pro	60.0%
AIME 2026 no tools	37.5%
LiveCodeBench v6	44.0%
Codeforces ELO	633
GPQA Diamond	43.4%
Tau2 (average over 3)	24.5%
BigBench Extra Hard	21.9%
MMMLU	67.4%
MMMU Pro	44.2%
OmniDocBench 1.5 (average edit distance, lower is better)	0.290
MATH-Vision	52.4%
MedXPertQA MM	23.5%
CoVoST	33.47
FLEURS (lower is better)	0.09
MRCR v2 8 needle 128k (average)	19.1%

Scores are Google's published results for the base google/gemma-4-E2B-it-assistant, not our own measurements. Quantization preserves the large majority of this; Q4_K_M and up stay close to full precision.

Choosing a quant

Quant	Size	Notes
Q4_K_S	78 MB	Compact 4-bit, fast.
Q4_K_M	78 MB	Recommended default. Best balance of size, speed and quality.
Q5_K_M	79 MB	Higher quality, low loss.
Q8_0	99 MB	Effectively lossless, reference quality.
F16	172 MB	Unquantized reference, twice the size of Q8_0.

Pick the largest file that fits your (V)RAM with room for context. Q4_K_M is the sweet spot for most setups; Q6_K or Q8_0 for maximum fidelity.

Get started

Run Gemma 4 E2B It Assistant locally with:

• Atomic Chat: the easiest path. Open the app, search AtomicChat/gemma-4-E2B-it-assistant-GGUF, pick a quant, hit Use this model.
• llama.cpp: llama-server -hf AtomicChat/gemma-4-E2B-it-assistant-GGUF:Q4_K_M --jinja -c 8192
• Ollama: ollama run hf.co/AtomicChat/gemma-4-E2B-it-assistant-GGUF:Q4_K_M
• LM Studio / Jan: search the repo id, download any quant.

Best practices

Parameter	Value
temperature	1.0
top_p	0.95
top_k	64

Google's recommended sampling configuration for google/gemma-4-E2B-it-assistant.

Run in llama.cpp

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

./llama.cpp/build/bin/llama-server \
    -hf AtomicChat/gemma-4-E2B-it-assistant-GGUF:Q4_K_M \
    --jinja -ngl 99 -c 8192 -fa on

How these were made

1. Download google/gemma-4-E2B-it-assistant (original weights).
2. Convert to f16 GGUF with llama.cpp.
3. Build an importance matrix over our calibration corpus.
4. Quantize the ladder with --imatrix.

License

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