all the models — AI benchmark observatory
← Models

Sakana Namazu

Sakana AI · proprietary · sakana-namazu

Sakana Namazu is Sakana AI's proprietary closed API model specialized for Japanese language and Japanese business contexts. It is a post-trained derivative of Moonshot's open Kimi K2.6 with in-house data for Japanese/business adaptation and reduced refusals/bias. Modalities: text, image, and file inputs → text output. Built-in tools include web_search and code_interpreter (code execution). Context window 256K. OpenAI-compatible API (model id sakana-namazu; pricing SKU sakana-namazu-v1.0). List pricing: $0.95 input / $0.15 cached input / $4.00 output per 1M tokens; thinking tokens billed as output. Tool fees (description only besides web_search field): web search $7.00 per 1,000 calls; code execution $0.12 per hour (session retained). Not yet available in the EU/EEA, UK, or Switzerland pending GDPR and region-specific compliance. Chart scores with catalog ids: AIME 2026 0.9667, MMLU-Pro 0.9033, LiveCodeBench v6 0.9033. No catalog id for JFBench 0.374, FairPoliticsQA 0.563, or in-house Translation 0.522 — recorded here only. Source: https://sakana.ai/namazu/

Not enough scores for a fingerprint yet.

Benchmark scores

BenchmarkScore
AIME 20260.97
MMLU-Pro0.90

Pricing

  • Sakana AI$0.95 / $4.00

Input / output per 1M tokens

AA metrics

No Artificial Analysis link yet.

Arena Elo

  • No Arena snapshot linked.
Sakana Namazu Benchmarks · all the models