Fugu Max is the cost-performance model in Sakana AI's Fugu family. Like Fugu Ultra it is not a monolithic language model but a learned multi-agent orchestration system: a model trained to route tasks across a swappable pool of underlying models and to recursively call instances of itself.
It keeps the family's 1,000,000-token context window and accepts text, image and file input, trading some of Ultra's peak reasoning depth for a substantially lower price per token.
Multi-agent orchestration at a price point that suits routine pull-request review and refactoring rather than only high-stakes work.
Lower per-token cost makes it viable for user-facing assistants where Ultra's pricing would not scale.
The 1,000,000-token window plus file input covers large report and codebase ingestion in a single request.
Tools, web search and structured output at roughly one fifth of Ultra's output price per token.
Released September 11, 2026.
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
Fugu Max is documented for high-volume coding workloads, where cost per task matters and it targets coding-related use cases. It is also documented for document extraction, especially when processing many documents efficiently. Another stated use case is multi-step automation, where tasks can be routed through a cost-efficient orchestration stack. It is further described as suitable for terminal and automation agents.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "sakana/fugu-max", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "sakana/fugu-max", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"sakana/fugu-max","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Fugu Max This page | Multi-agent orchestration at scale | |||
| Reasoning + agents | ||||
| Deep reasoning + long-context agents | ||||
| Reasoning + agents | ||||
| Long-context, multimodal & agentic workflows |
Fugu Max is the cost-performance model in Sakana AI's Fugu family. Like Fugu Ultra it is a learned multi-agent orchestration system rather than a monolithic language model, trained to route tasks across a swappable pool of underlying models and to recursively call instances of itself.
It keeps the family's 1,000,000-token context window and the same input types, trading some of Ultra's peak reasoning depth for a substantially lower price per token.
Sakana AI.
1,000,000 tokens, with up to 128,000 tokens of output.
It takes image, text and file input and returns text.
Yes. It supports function calling, structured outputs, streaming, reasoning and web search.
Everyday coding and review, user-facing assistants, long-document processing and high-volume agentic pipelines where Ultra's pricing would not scale.
Send a request to the chat completions endpoint with the model id sakana/fugu-max. The model also answers to the alias fugu-max.
