Skip to main content

Shisa LLM

Shisa LLM is a family of large language models tuned for Japanese, fluent in English, and served behind an OpenAI-compatible API. If you already use the OpenAI SDKs, you can point them at Shisa by changing two lines — the base URL and the model name.

POST https://api.shisa.ai/openai/v1/chat/completions

Why Shisa LLM

  • Japanese expertise — trained and evaluated on Japanese data, with strong cultural and linguistic nuance rather than translated English behaviour.
  • Bilingual proficiency — high quality across Japanese and English, and reliable switching between them in a single conversation.
  • OpenAI-compatible — works with openai-python, openai-node, and any client that speaks the Chat Completions API.
  • Streaming — token-by-token Server-Sent Events for responsive UIs.
  • Japan-hosted — inference runs in Japan for data-residency-sensitive workloads.
  • Reduced language leakage — markedly fewer stray non-Japanese tokens than general-purpose models (up to 47.8× reduction versus the base model).

Benchmarks

MetricScoreModel
Japanese MT-Bench9.26Shisa V2.1 Pro
Japanese average73.1%Shisa V2.1 Pro
Japanese average80.5%Shisa V2 405B
Language-leakage reduction47.8×vs. base model

Read the technical write-ups for Shisa V2.1 and Shisa V2 405B.

Next steps