Model Detail
gemma-4-31B-it-GGUF
—gemma-4-31B-it-GGUF is a multimodal model with 31B parameters released by unsloth. The model is registered under the image-text-to-text pipeline tag on Hugging Face, and supports text+image+video->text inputs, distributed under the permissive apache-2.0 license.
gemma-4-31B-it-GGUF is priced at $0.13/M input tokens and $0.38/M output tokens. Operationally the model offers a 262K-token context window, which matters when sizing it for prompt-heavy or latency-sensitive workloads. At this input rate the model sits in the commodity tier and is suitable for high-volume workloads where per-call cost dominates the decision.
gemma-4-31B-it-GGUF ships with 31B parameters, distributed as a quantized weight variant for lower-VRAM inference. The apache-2.0 license is permissive, allowing commercial deployment and derivative work without per-seat fees, though attribution requirements still apply.
Downloads of gemma-4-31B-it-GGUF have moved -15.9% over the trailing seven days, -39.8% over the trailing thirty days. That is a slight downtrend, consistent with normal cooling as newer models compete for the same workloads. These numbers are signal, not guarantee — week-over-week download counts on Hugging Face also reflect mirror traffic, CI scrapes, and one-off benchmarking runs.
gemma-4-31B-it-GGUF is best fit for mixed text-and-image reasoning tasks such as document understanding, high-volume batch jobs where per-call cost dominates the budget, and long-context tasks such as full-codebase analysis or book-length summarization (262K tokens). 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.
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
Borrowed Geometry: Cross-Distribution Head-Importance Fingerprints of Frozen Pretrained Gemma 4 31B
arXiv:2605.00333v2 Announce Type: replace-cross Abstract: Frozen Gemma 4 31B weights pretrained exclusively on text, unmodified, transfer through a thin trainable interface to non-text modalities the substrate has never processed. On the L24--L29 slice (192 attention heads), an English-text TxtCopy
PSK at SemEval-2026 Task 9: Multilingual Polarization Detection Using Ensemble Gemma Models with Synthetic Data Augmentation
arXiv:2605.05159v1 Announce Type: new Abstract: We present our system for SemEval-2026 Task 9: Multilingual Polarization Detection, a binary classification task spanning 22 languages. Our approach fine-tunes separate Gemma~3 models (12B and 27B parameters) per language using Low-Rank Adaptation (LoR
MedGemma 1.5 Technical Report
arXiv:2604.05081v2 Announce Type: replace Abstract: We introduce MedGemma 1.5 4B, the latest model in the MedGemma collection. MedGemma 1.5 expands on MedGemma 1 by integrating additional capabilities: high-dimensional medical imaging (CT/MRI volumes and histopathology whole slide images), anatomica
Distilling Self-Consistency into Verbal Confidence: A Pre-Registered Negative Result and Post-Hoc Rescue on Gemma 3 4B
arXiv:2604.24070v1 Announce Type: cross Abstract: Small instruct-tuned LLMs produce degenerate verbal confidence under minimal elicitation: ceiling rates above 95%, near-chance Type-2 AUROC, and Invalid validity profiles. We test whether confidence-conditioned supervised fine-tuning (CSFT) with self
