Sakana Namazu

1.65
What caught my attention about Namazu wasn’t the “Japanese LLM” label. It was what Sakana AI actually changed. The model is built on Kimi K2.6. Sakana didn’t try to build a new foundation model from scratch. Instead, it focused on Japanese post-training, tool use, and packaging the whole thing into an API that developers can actually plug into a product.
Advertisement 728 × 90
CompanySakana
Context262K
Released2026-08
Updated2026-08-21

Sakana Namazu Overview

Sakana Namazu is a version of Moonshot’s Kimi K2.6 that has been post-trained by Sakana AI for Japanese language use and Japanese cultural context.

The base model uses a Mixture-of-Experts architecture with roughly 1 trillion total parameters and about 32 billion active parameters during inference.

It covers the usual API basics:

Text input
Image input
Text output
Function calling
Web search
Code execution

The last two are the more interesting additions.

Namazu can decide when it needs to search the web or run code instead of waiting for the developer to trigger each step manually. It can also use those tools several times inside the same task.

Sakana has shown a weekly market-research example where the model breaks the task down, searches for information, checks sources against each other, and then writes the report.

The API is OpenAI-compatible. If your app already uses an OpenAI-style SDK, switching usually comes down to changing the base_url and API key.

Sakana Namazu Pricing

PlanPriceDescription
Free version Free Available on OpenRouter as sakana/sakana-namazu:free
Input tokens $0.95 / 1M Standard input
Cached input tokens $0.15 / 1M Charged when cache is hit
Output tokens $4.00 / 1M Thinking tokens are billed at the same output rate
Web search $7.00 / 1,000 calls Includes text retrieval
Code execution $0.12 / hour Billing continues while the session is kept active

Pricing is based on information published by Sakana AI and OpenRouter. Standard billing is in US dollars, while enterprise plans can support billing in Japanese yen.

There is one thing to check before using the free OpenRouter version with real data.

Sakana may use submitted inputs for model training unless you opt out in the Console. Data processing is also not guaranteed to stay entirely within Japan.

  • There is a free version: useful for checking Japanese output quality and tool behavior before paying.
  • The main API is usage-based: no monthly commitment; input, output, search, and code execution are billed separately.
  • For light testing: the free OpenRouter option should be enough to get a feel for the model.
  • For production: the paid API makes more sense once usage becomes regular.
  • For companies: review the training-data and processing-location terms before sending customer or internal data.

Sakana Namazu Key Features

1. The Japanese tuning goes beyond grammar

Kimi K2.6 can already handle Japanese, so Namazu wouldn’t be especially interesting if all it did was make sentences sound a little smoother.

Sakana’s work goes further into business phrasing, honorific language, and how the model handles politically or historically sensitive topics.

The company published results from an internal benchmark called FairPoliticsQA, which measures neutrality in responses about political and historical issues.

Kimi K2.6 scored 34.10%. Namazu scored 56.30%.

I wouldn’t read too much into that number on its own. It’s a Sakana benchmark, and one neutrality test obviously doesn’t represent overall Japanese-language quality.

What it does show is where the tuning is aimed. Sakana isn’t only changing vocabulary and fluency. It’s also trying to adjust how the model responds inside a Japanese social and business context.

You may not notice much of a difference in casual chat. Customer emails, support replies, brand content, and formal business writing are much more useful tests.

2. Search and code execution are built in

Namazu comes with web search and code execution as built-in tools.

If a task needs fresh information, the model can search. If it needs calculations, data processing, or code validation, it can run code.

It can move between those tools several times without requiring the user to manually break the job into separate prompts.

Take market research as an example. A more basic workflow might involve asking a model for search terms, doing the search yourself, pasting the results back in, and then asking for analysis.

Namazu can handle more of that chain on its own.

None of this is a brand-new Agent concept. The practical benefit is simpler: developers have less plumbing to build themselves.

3. Japanese tuning without a higher token price

Namazu uses an OpenAI-compatible API, so it fits fairly easily into existing applications.

Input costs $0.95 per million tokens and output costs $4.00 per million tokens, matching Kimi K2.6 pricing.

That’s probably the cleanest part of the pitch.

Sakana added another layer of Japanese-focused training without adding another layer to the token price.

If you’re already choosing between Kimi K2.6 and other models in the same price range, and Japan is one of your main markets, Namazu is easy to justify testing.

I’d spend less time staring at benchmark tables and more time running the same real Japanese prompts through both models.

4. The earlier work felt like research. Namazu feels like a product.

Before Namazu, Sakana had already experimented with post-training models such as DeepSeek and Llama for Japanese language behavior and response neutrality.

Those projects felt much closer to research prototypes.

Namazu changes the setup.

The base is now Kimi K2.6, search and code execution are available through the API, and developers don’t have to download weights, deploy their own inference stack, and then bolt on a separate tool system.

The earlier experiments were asking whether this type of post-training could work.

Namazu is closer to asking whether people will actually use it in production.

That’s a much more useful test.

Summary

Strengths

  • Tuned for Japanese business writing, honorific language, and local context
  • Token pricing stays in line with Kimi K2.6
  • Built-in web search and code execution
  • Can handle multi-step tasks involving search, checking, and computation
  • OpenAI-compatible API makes migration relatively easy
  • Free version available on OpenRouter for testing

Limitations

  • Sakana has not clearly published the official context-window size; third-party listings show 262K
  • Independent benchmark coverage is still limited
  • Input data may be used for training unless you opt out
  • Data processing is not guaranteed to remain entirely within Japan
  • The service is not available in the EU, UK, or Switzerland

Who It’s For

Namazu makes the most sense for teams building for the Japanese market and working with a lot of Japanese-language content.

It’s especially worth testing for:

  • Business writing
  • Customer support
  • Research workflows
  • Japanese-language Agents
  • Existing Kimi K2.6 users who want better Japanese handling

If you already use Kimi K2.6, the easiest way to judge Namazu is simply to run the same production prompts through both and compare the results.

Who Should Skip It

Namazu is a poor fit if your company requires all business data to stay inside Japan with no exceptions.

It also won’t suit projects that require context windows above 1M tokens, or teams that only deploy fully open-source, self-hosted models.

And if Japanese barely matters to your product, there isn’t much reason to switch just for the sake of using Namazu.

Comments (0)

Leave a comment

Advertisement 728 × 90

Sakana Model Comparison

Model Context Pricing API Released Global Heat
Sakana Namazu
262K YES 2026-08
33/100
1M YES 2026-06
38/100

Similar Models