Skip to content
← Models
Open weightsKARAKURI Inc./10 trustMetrics-checked

KARAKURI LM 8x7B Instruct v0.1

Japanese-English MoE chat model with function calling and RAG support.

Visit official page →

About

KARAKURI LM 8x7B Instruct v0.1 is a mixture-of-experts language model developed by KARAKURI Inc. for English and Japanese. It is a fine-tuned version of tokyotech-llm/Swallow-MX-8x7b-NVE-v0.1, trained on approximately 1 billion tokens of fine-tuning data. The vendor describes it as the first domestic Japanese LLM with function calling and RAG support.

The model is licensed under Apache 2.0 and is distributed via Hugging Face and AWS Marketplace. Development used AWS Trainium, which the vendor says cut training cost by up to 50%. It is positioned for business use cases in Japanese enterprise settings, with evaluation across helpfulness, correctness, coherence, and creativity dimensions.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

June 1, 2024

Benchmarks

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

Use cases

japanese-nlpchatfunction-callingrag

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 128; Trending score 72/100 from researcher assessment.

Sources