What are the hardware requirements for Gemma 4 31B 5bit?

Gemma 4 31B 5bit requires approximately 22.3 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Gemma 4 31B 5bit and how large is it?

Gemma 4 31B 5bit is developed by mlx-community, a model with 31b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Gemma 4 31B 5bit perform on standard benchmarks?

Gemma 4 31B 5bit has the following published scores: LMSYS ELO 1451. Compare against reference models on this page or on the LLM Explorer leaderboards.

Gemma 4 31B 5bit by mlx-community — VRAM 22.3GB

Name: Gemma 4 31B 5bit
Author: mlx-community

Gemma 4 31B 5bit is an open-source language model by mlx-community. Features: 31b LLM, VRAM: 22.3GB, License: apache-2.0, Quantized, LLM Explorer Score: 0.38, ELO: 1451.

5-bit 5bit Gemma4 Image-text-to-text Mlx Quantized Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/mlx-community/gemma-4-31b-5bit

Gemma 4 31B 5bit Benchmarks

LMSys ELO 1451 vs 1272 (so35) 14.1%

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

Gemma 4 31B 5bit Parameters and Internals

LLM Name	Gemma 4 31B 5bit
Repository 🤗	https://huggingface.co/mlx-community/gemma-4-31b-5bit
Base Model(s)	google/gemma-4-31b google/gemma-4-31b
Model Size	31b
Required VRAM	22.3 GB
Updated	2026-08-08
Maintainer	mlx-community
Model Type	gemma4
Model Files	5.4 GB: 1-of-5 5.3 GB: 2-of-5 5.3 GB: 3-of-5 5.4 GB: 4-of-5 0.9 GB: 5-of-5
Quantization Type	5bit
Model Architecture	Gemma4ForConditionalGeneration
License	apache-2.0
Transformers Version	5.5.0.dev0
Tokenizer Class	GemmaTokenizer
Padding Token	<pad>

Rank the Gemma 4 31B 5bit capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Gemma 4 31B 5bit

Best Alternatives	Context / RAM	Downloads	Likes
...emma 4 31B It Unsloth Bnb 4bit	0K / 19.2 GB	313172	22
Gemma 4 31B It 4bit	0K / 18.5 GB	45067	46
... 4 31B It Qat Q4 0 Unquantized	0K / 62.6 GB	9972	39
Gemma 4 31B It 8bit	0K / 33.8 GB	3214	17
Gemma 4 31B 8bit	0K / 33.8 GB	957	24
Gemma 4 31B It 6bit	0K / 26 GB	906	7
Gemma 4 31B It Qat W4a16 Ct	0K / 23.3 GB	2187795	55
Gemma 4 31B 4bit	0K / 18.5 GB	448	11
Gemma 4 31B 6bit	0K / 26 GB	234	2
Gemma 4 31B It 5bit	0K / 22.3 GB	159	3

Note: green Score (e.g. "73.2") means that the model is better than mlx-community/gemma-4-31b-5bit.