name = "Sakana Namazu"
description = "Japanese-specialized reasoning model based on Kimi K2.6 and tuned for Japanese language, culture, and business workflows"
family = "sakana-namazu"
release_date = "2026-08-03"
last_updated = "2026-08-03"
attachment = true
reasoning = true
temperature = true
tool_call = true
structured_output = true
open_weights = false

[limit]
context = 262_144
output = 65_536

[modalities]
input = ["text", "image", "pdf"]
output = ["text"]

[[links]]
label = "Official product page"
url = "https://sakana.ai/namazu/"
type = "announcement"

[[links]]
label = "Official model documentation"
url = "https://console.sakana.ai/models?model=sakana-namazu"
type = "docs"

[[benchmarks]]
name = "AIME26"
score = 96.67
source = "https://console.sakana.ai/models?model=sakana-namazu"

[[benchmarks]]
name = "MMLU-Pro"
score = 90.33
source = "https://console.sakana.ai/models?model=sakana-namazu"

[[benchmarks]]
name = "LiveCodeBench v6"
score = 90.33
source = "https://console.sakana.ai/models?model=sakana-namazu"

[[benchmarks]]
name = "JFBench"
score = 37.40
source = "https://console.sakana.ai/models?model=sakana-namazu"

[[benchmarks]]
name = "Translation"
score = 52.20
source = "https://console.sakana.ai/models?model=sakana-namazu"

[[benchmarks]]
name = "FairPoliticsQA"
score = 56.30
source = "https://console.sakana.ai/models?model=sakana-namazu"