BestLLMfor EN Your hardware. Your LLM. Your call.
APIOpen data Find my LLM
Model fiche

Rakuten AI 3.0

By Rakuten · Japan

Updated 2026-07-13

chat general multilingual moe
Parameters
700B
License
Apache 2.0
Context
32k
VRAM (Q4)
420 GB
Released
March 2025

Overview

Rakuten's flagship ~700B MoE model built under Japan's GENIAC program and released under Apache 2.0. Best-in-class Japanese performance with serious enterprise e-commerce DNA.

When to pick this model

  • Japanese-language production workloads (support, content, search)
  • Bilingual JP/EN e-commerce and retail applications
  • Enterprise deployments needing Apache 2.0 licensing
  • Replacing closed JP models like proprietary Rakuten/LINE APIs
  • Localization pipelines targeting the Japanese market

VRAM requirements by quantization

VRAM REQUIRED (GB)128256512Q4_K_M420 GBQ5_K_M500 GBQ8_0745 GBFP161400 GB
QuantizationVRAM required
Q4_K_M (recommended)420 GB
Q5_K_M500 GB
Q8_0745 GB
FP16 (no quantization)1400 GB

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

In practice, Rakuten AI 3.0 is server-class even at Q4_K_M (420 GB). Stepping up to Q8_0 nearly doubles the footprint to 745 GB, and unquantized FP16 weights take 1400 GB — plan your GPU around the Q4 or Q5 figure unless you specifically need the higher fidelity.

Without a GPU, Rakuten AI 3.0 needs roughly 500 GB of system RAM to run on CPU via llama.cpp or Ollama — workable for background jobs, but far slower than GPU inference. Throughput estimates from our compatibility engine: around 1 tokens/sec on entry-level GPUs, on the order of 4 tokens/sec on a mid-range card, and up to 12 tokens/sec on high-end hardware — assuming the chosen quantization fully fits in VRAM.

What hardware do you need

The table below matches Rakuten AI 3.0 to common GPU memory tiers, using the highest-fidelity quantization that fully fits each card class. Spilling layers to system RAM works but costs most of the speed, so size your card to the quantization you actually want to run.

GPU memoryExample cardsBest fit for Rakuten AI 3.0
8 GBRTX 5070 Laptop, RTX 5060, RTX 5060 Ti 8GBDoes not fit — needs 420 GB at Q4_K_M
12 GBRTX 5070, RTX 5070 Ti Laptop, RTX 4080 LaptopDoes not fit — needs 420 GB at Q4_K_M
16 GBRTX 5080, RTX 4080 Super, Radeon RX 9070 XTDoes not fit — needs 420 GB at Q4_K_M
24 GBRTX 4090, Radeon RX 7900 XTX, RTX 5090 LaptopDoes not fit — needs 420 GB at Q4_K_M
32 GBRTX 5090Does not fit — needs 420 GB at Q4_K_M

Which GPU should you buy to run Rakuten AI 3.0?

To run Rakuten AI 3.0 locally at Q4, you need ~420 GB of VRAM. The best value for this is a Apple Mac Studio (64+ GB unified memory).

Check Apple Mac Studio price on Amazon →

As an Amazon Associate, BestLLMfor earns from qualifying purchases, at no extra cost to you. It does not influence our independent rankings.

Strengths

  • Top-tier Japanese fluency, beating most open models on JP benchmarks
  • 700B total parameters give it broad knowledge depth
  • Apache 2.0 — no commercial restrictions
  • Backed by Rakuten's massive e-commerce and fintech corpus
  • Built under Japan's GENIAC sovereign-AI initiative

Limitations

  • Roughly 420 GB VRAM in Q4 — datacenter-only
  • 32k context is tight versus modern 128k+ flagships
  • Heavily skewed toward Japanese and commerce; weaker on global general tasks

Typical workloads

In our catalog grid, Rakuten AI 3.0 is filed under Native Japanese, JP Sovereignty — the use cases where its size/quality trade-off makes the most sense. Its tags translate to concrete workloads: multilingual workloads.

The 32k-token context window covers long chats and mid-sized documents, though very large retrieval workloads will need chunking. The Apache 2.0 license is permissive, so shipping it inside a commercial product raises no special legal questions.

Architecture & training

Architecture: MoE · 700B total · Rakuten AI 3 · 32k context

Training: Rakuten — massive JP/EN corpus for e-commerce and enterprise.

Verdict

The default open-weight choice for Japanese enterprise; overkill and underspecialized for anyone else.

Quick start

# Infrastructure lourde requise — non disponible en local standard

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.

Similar models worth comparing

Frequently asked questions

How much VRAM does Rakuten AI 3.0 need?

At the recommended Q4_K_M quantization, Rakuten AI 3.0 needs about 420 GB of VRAM. Q8_0 takes 745 GB, and unquantized FP16 weights take 1400 GB.

Can Rakuten AI 3.0 run without a GPU?

Yes — with roughly 500 GB of system RAM it runs CPU-only through llama.cpp or Ollama. Expect a fraction of GPU speed, which is fine for background or batch jobs but slow for interactive chat.

What context window does Rakuten AI 3.0 support?

Rakuten AI 3.0 supports a 32k-token context window (32,768 tokens).

Can I use Rakuten AI 3.0 commercially?

Yes. Rakuten AI 3.0 is released under Apache 2.0, a permissive open-source license that allows commercial use, modification and redistribution.

How fast is Rakuten AI 3.0 on consumer hardware?

Our compatibility engine estimates on the order of 4 tokens/sec on a mid-range GPU and up to 12 tokens/sec on high-end cards, assuming the quantization fully fits in VRAM.

Which quantization of Rakuten AI 3.0 should I download first?

Start with Q4_K_M (420 GB) — the standard size/quality sweet spot. Step up to Q5_K_M or Q8_0 only if you have VRAM headroom. It does not fit a single 24 GB consumer card — plan for multi-GPU or server hardware.

Tools

Is Rakuten AI 3.0 the right pick for you?

Compute self-hosted ROI → Back to catalog