Base model: google/gemma-4-E4B-it
Gemma 4 E4B, self-quantized to MLX 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.
- 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 MLXs are self-quantized from the original weights, not a repack. The importance matrix keeps low-bit quants closer to the full-precision model.
Model Overview
| Property | Value |
|---|---|
| Base model | google/gemma-4-E4B-it |
| Parameters | 4.5B effective (8B with embeddings) |
| Layers | 42 |
| Sliding window | 512 tokens |
| Context length | 128K tokens |
| Vocabulary | 262K |
| Modalities | Text, Image, Audio |
| Architecture | Dense decoder, hybrid sliding-window (512) and global attention, 8 attention heads over 2 KV heads, Gemma4ForConditionalGeneration |
| This repo | MLX weights |
Benchmarks
| Benchmark | Score |
|---|---|
| MMLU Pro | 69.4% |
| AIME 2026 no tools | 42.5% |
| LiveCodeBench v6 | 52.0% |
| Codeforces ELO | 940 |
| GPQA Diamond | 58.6% |
| Tau2 (average over 3) | 42.2% |
| BigBench Extra Hard | 33.1% |
| MMMLU | 76.6% |
| MMMU Pro | 52.6% |
| OmniDocBench 1.5 (average edit distance, lower is better) | 0.181 |
| MATH-Vision | 59.5% |
| MedXPertQA MM | 28.7% |
| CoVoST | 35.54 |
| FLEURS (lower is better) | 0.08 |
| MRCR v2 8 needle 128k (average) | 25.4% |
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.
Get started
- Atomic Chat: search
AtomicChat/gemma-4-E4B-it-MLX-8bitand hit Use this model. - mlx-lm:
mlx_lm.generate --model AtomicChat/gemma-4-E4B-it-MLX-8bit --prompt "Hello" --max-tokens 512 - Server:
mlx_lm.server --model AtomicChat/gemma-4-E4B-it-MLX-8bit --port 8080
Best practices
| Parameter | Value |
|---|---|
| temperature | 1.0 |
| top_p | 0.95 |
| top_k | 64 |
Google's recommended sampling configuration for google/gemma-4-E4B-it.
How these were made
1. Download google/gemma-4-E4B-it (original weights).
2. Convert and quantize with mlx_lm.convert on our pipeline.
License
Original model by Google, released under the Apache 2.0 license. Full terms: Apache 2.0. Quantized by Atomic Chat.