Models·Sakana

Sakana Namazu: Japanese Reasoning Model

sakana/sakana-namazu

Sakana Namazu is a Japanese-focused reasoning model with image and file input, tool support, and text output.

$0.95Input, per million tokens
$4Output, per million tokens
262KContext window
66KLongest answer

262K tokens of context — about 196,608 words, on a scale that ends at 2M.

What it can do

  • Reasoning
  • Vision
  • Tool use
  • Takes text
  • Takes image
  • Takes file
  • Gives text

Released 11 Aug 2026 · checked 3 Oct 2026.

About it

Sakana Namazu is a Japanese-specialized reasoning model from Sakana, based on Kimi K2.6 and trained further for Japanese language and business contexts. It accepts text, images, and files, can use tools, and returns text. The model has a 262,144-token context window and costs $0.95 per million input tokens and $4 per million output tokens.

Review

In a line

Sakana Namazu is Sakana’s reasoning model for Japanese instructions and business-related work. It is based on Kimi K2.6 with extra training for Japanese language and business contexts.

The model accepts text, images, and files, and produces text answers. It was released on August 11, 2026.

What it's good at

Its main fit is Japanese instruction following and reasoning. The model is also designed for business contexts.

Namazu supports image input and tool use, so tasks can combine written instructions with visual or file-based material.

Price in plain terms

Input costs $0.95 per million tokens, while output costs $4 per million tokens. In simple terms, reading uses the lower rate and generated answers use the higher rate.

At the stated rough usage, reading a 500-word page costs about 0.062 cents. One chat turn with about 1,000 input tokens and 500 output tokens costs about 0.29 cents; 1,000 such turns cost about $2.95.

Context and limits

Namazu can take up to 262,144 tokens in one context, which is the material it can consider together. A single answer can contain up to 65,536 tokens.

Its output is text only. The listed inputs are text, images, and files, and its reasoning and tool support do not change that output format.

Ways to use it

Use Namazu in an app or chat tool that offers the model, including workflows that pass in Japanese text, images, or files. Its tool support can let those workflows call supported tools.

For model selection, the identifier is `sakana/sakana-namazu`. You can compare it with other entries in the Sakana models list or browse vision language models.

Verdict

Choose Namazu when Japanese instruction following, reasoning, and business-oriented language work matter, especially if image or file input is useful. Its large context can hold substantial material, and the rough chat cost is low for the stated example.

Look elsewhere if you need non-text answers or a documented integration list. The facts here describe tool support and accepted inputs, but not specific apps, providers, or integrations.

Verdict

Sakana Namazu suits readers who need Japanese-focused reasoning with text, image, or file inputs. Its long context is useful for supplying extensive material, while the example costs stay below three dollars for 1,000 chat turns. The model returns text only, so it is not a fit for projects that require image or audio output. Buyers should also confirm that their chosen app or service exposes this model and its tool features before committing.

  • Japanese-focused training for language and business contexts
  • Reasoning and tool support
  • Accepts text, images, and files
  • 262,144-token context window
  • Low example cost for 1,000 chat turns
  • Text output only
  • No specific integrations are listed
  • Output costs more than input at $4 per million tokens
  • Pick it ifJapanese instruction following and business tasks; Workflows using images or files; Long-context text reasoning
  • Skip it ifYou need image or audio output; You require named integrations or providers; Your work does not involve Japanese-focused tasks

Pricing

Input is $0.95 per million tokens and output is $4 per million tokens. The rough examples are about 0.062 cents for one page, 0.29 cents for one chat turn, and $2.95 for 1,000 such turns.

Questions

What is Sakana Namazu?

It is Sakana’s Japanese-specialized reasoning model, based on Kimi K2.6 with additional training for Japanese language and business contexts.

What can I send to Namazu?

The model accepts text, images, and files.

Can Namazu use tools?

Yes. Tool support is listed for this model.

How much does Namazu cost?

Input costs $0.95 per million tokens and output costs $4 per million tokens.

How much text can it consider at once?

Its context window is 262,144 tokens.

How long can one answer be?

The maximum output is 65,536 tokens, and the output modality is text.

What is the model identifier?

Its model ID is `sakana/sakana-namazu`.

On these boards

More from Sakana