Skip to content
← Models
Open weightsRakuten Group/10 trustMetrics-checked

RakutenAI-7B

Apache 2.0 Japanese LLM built on Mistral 7B by Rakuten.

Visit official page →

About

RakutenAI-7B is a Japanese-oriented large language model released by Rakuten Group in March 2024, retrofitting Mistral 7B v0.1 for Japanese. The key change is a tokenizer extended from 32K to 48K tokens with 16,000 added Japanese tokens, which improves character-per-token rates and stretches effective context. The team applied rigorous data filtering, normalization and deduplication over Japanese and English corpora, and released foundation, instruct and chat variants under the Apache 2.0 license. At release it achieved the best scores among open 7B models on Japanese language-understanding benchmarks while keeping English performance competitive.

Rakuten later built on it with Rakuten AI 2.0, a Mistral-based MoE line released in 2025. The 7B model is a solid, genuinely open foundation for Japanese applications, though it is now several generations behind the frontier and was superseded by larger successors in Rakuten's own lineup.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

March 21, 2024

Benchmarks

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

japanese-nlpchatfine-tuning

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 91; Trending score 72/100 from researcher assessment.

More in the rakutenai family

Sources