Model Detail
gemma-4-26b-a4b-it-UD-MLX-4bit
—gemma-4-26b-a4b-it-UD-MLX-4bit is a multimodal model with 26B parameters released by unsloth. The model is registered under the image-text-to-text pipeline tag on Hugging Face, distributed under the permissive apache-2.0 license.
gemma-4-26b-a4b-it-UD-MLX-4bit ships with 26B parameters. Total weight footprint is approximately 25.2 GB, which is the relevant figure when planning local-inference VRAM. The apache-2.0 license is permissive, allowing commercial deployment and derivative work without per-seat fees, though attribution requirements still apply.
gemma-4-26b-a4b-it-UD-MLX-4bit is best fit for mixed text-and-image reasoning tasks such as document understanding. Treat this as a starting matrix rather than a benchmark verdict — the right deployment usually depends on the specific evaluation suite that mirrors your workload.
Words Speak Louder Than Order: A Behavioral Evaluation of Gemma 4
arXiv:2609.30716v1 Announce Type: cross Abstract: When a language model receives two conflicting documents as input, how does it decide which one to prioritize? Does it rely on how the sources are framed or the presentation order of the documents? We evaluated this behavior on Google's pre-trained G
Recurrence Is Not Enough: Causally Validating Multilingual SAE Translation Features in Gemma 2 and 3
arXiv:2609.04808v1 Announce Type: cross Abstract: Sparse autoencoder (SAE) features are increasingly used to explain and steer language-model behavior, but it remains unclear whether a feature found in one language context plays the same causal role when processing prompts in another language. We st
Gemma 4 Technical Report
arXiv:2607.02770v2 Announce Type: replace Abstract: We introduce Gemma 4, a new generation of open-weight, natively multimodal language models in the Gemma model family. Designed to advance compute efficiency and reasoning, the Gemma 4 model suite features dense and Mixture-of-Experts architectures,
Do Active SAE Feature Planes Carry More Holonomy? A Preregistered Reversal in Gemma
arXiv:2607.20522v1 Announce Type: new Abstract: This paper tests whether holonomy concentrates on active sparse-autoencoder (SAE) feature planes in Gemma 2 2B, a concrete operationalization of the broader semantic-concentration prediction. Holonomy is measured at the final-token layer-12 to layer-13
Hugging Face and Cerebras bring Gemma 4 to real-time voice AI
Fine-Tuning and Serving Gemma 4 31B on Google Cloud TPU: A Technical Comparison with GPU Baselines
arXiv:2605.25645v2 Announce Type: replace-cross Abstract: We present the first end-to-end demonstration of fine-tuning and serving Google's Gemma 4 31B model on TPU hardware, providing an empirical comparison of TPU and GPU platforms for large language model adaptation. Using LoRA on a Google TPU v5
