What are the hardware requirements for RakutenAI 3.0 MLX 4bit?

RakutenAI 3.0 MLX 4bit requires approximately 198.7 GB of VRAM and supports a context window of 160K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed RakutenAI 3.0 MLX 4bit and how large is it?

RakutenAI 3.0 MLX 4bit is developed by mlx-community, a model with 671b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate RakutenAI 3.0 MLX 4bit?

RakutenAI 3.0 MLX 4bit is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

RakutenAI 3.0 MLX 4bit by mlx-community — VRAM 198.7GB, 160K context

Name: RakutenAI 3.0 MLX 4bit
Author: mlx-community

RakutenAI 3.0 MLX 4bit is an open-source language model by mlx-community. Features: 671b LLM, VRAM: 198.7GB, Context: 160K, License: apache-2.0, Quantized, LLM Explorer Score: 0.23.

4-bit 4bit Base model:quantized:rakuten/r... Base model:rakuten/rakutenai-3... Conversational Deepseek v3 En Ja Mlx Quantized Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/mlx-community/RakutenAI-3.0-MLX-4bit

RakutenAI 3.0 MLX 4bit Parameters and Internals

LLM Name	RakutenAI 3.0 MLX 4bit
Repository 🤗	https://huggingface.co/mlx-community/RakutenAI-3.0-MLX-4bit
Base Model(s)	Rakuten/RakutenAI-3.0 Rakuten/RakutenAI-3.0
Model Size	671b
Required VRAM	198.7 GB
Updated	2026-07-25
Maintainer	mlx-community
Model Type	deepseek_v3
Model Files	3.7 GB: 1-of-88 4.4 GB: 2-of-88 4.2 GB: 3-of-88 4.4 GB: 4-of-88 4.4 GB: 5-of-88 4.2 GB: 6-of-88 4.4 GB: 7-of-88 4.4 GB: 8-of-88 4.2 GB: 9-of-88 4.4 GB: 10-of-88 4.4 GB: 11-of-88 4.2 GB: 12-of-88 4.4 GB: 13-of-88 4.4 GB: 14-of-88 4.2 GB: 15-of-88 4.4 GB: 16-of-88 4.4 GB: 17-of-88 4.2 GB: 18-of-88 4.4 GB: 19-of-88 4.4 GB: 20-of-88 4.2 GB: 21-of-88 4.4 GB: 22-of-88 4.4 GB: 23-of-88 4.2 GB: 24-of-88 4.4 GB: 25-of-88 4.4 GB: 26-of-88 4.2 GB: 27-of-88 4.4 GB: 28-of-88 4.4 GB: 29-of-88 4.2 GB: 30-of-88 4.4 GB: 31-of-88 4.4 GB: 32-of-88 4.2 GB: 33-of-88 4.4 GB: 34-of-88 4.4 GB: 35-of-88 4.2 GB: 36-of-88 4.4 GB: 37-of-88 4.4 GB: 38-of-88 4.2 GB: 39-of-88 4.4 GB: 40-of-88 4.4 GB: 41-of-88 4.2 GB: 42-of-88 4.4 GB: 43-of-88 4.4 GB: 44-of-88 4.2 GB: 45-of-88 4.4 GB: 46-of-88
Supported Languages	en ja
Quantization Type	4bit
Model Architecture	DeepseekV3ForCausalLM
License	apache-2.0
Context Length	163840
Model Max Length	163840
Transformers Version	4.57.1
Vocabulary Size	129280
Torch Data Type	bfloat16

Rank the RakutenAI 3.0 MLX 4bit 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 RakutenAI 3.0 MLX 4bit

Best Alternatives	Context / RAM	Downloads	Likes
DeepSeek V3.1 Terminus 4bit	160K / 190.1 GB	42	2
DeepSeek V3.1 8bit	160K / 177.8 GB	4268	3
DeepSeek V3.1 4bit	160K / 190.1 GB	2974	6
DeepSeek V3.1 Base 4bit	160K / 190.1 GB	43	3
DeepSeek R1 0528 5bit	160K / 231 GB	66	3
DeepSeek V3 AWQ	160K / 351.9 GB	247	36
R1 1776	160K / 243.4 GB	855	2324
DeepSeek Prover V2 671B	160K / 189 GB	562	831
Cogito 671B V2.1	160K / 219.9 GB	85	29
Deepseek V3.1 MLX 5.5bit	160K / 231 GB	339	5

Note: green Score (e.g. "73.2") means that the model is better than mlx-community/RakutenAI-3.0-MLX-4bit.