Fugu Max | EmpirioLabs AI Docs

Sakana AI · Text Generation

POST /v1/chat/completions

Cost-efficient multi-agent conductor that assembles a right-sized expert team per task, with 1M context, image input, and web search.

At a glance

Field	Value
Model id	`fugu-max`
Model release date	2026-09-10
Input modalities	Text, Image
Output modalities	Text
Context window	1M
Weight precision	-
Max output tokens	131,072
Features	reasoning, multimodal, web_search, function_calling, agentic_coding, structured_output, cache
Native inference	No
New	Yes
Structured output	JSON Schema
Supported endpoints	`POST /v1/chat/completions`, `POST /v1/responses`, `POST /v1/messages`, `POST /v1beta/models/fugu-max:generateContent`
Alternate model ids	`fugu-max-v1.0`, `sakana/fugu-max`, `sakana/fugu-max-v1.0`

Pricing

Charge	Spec	Rate
Input	per 1M prompt tokens	$2.00
Output	per 1M generated tokens	$6.00
Implicit cache read	per 1M cached input tokens	$0.25
Web search	per call when invoked	$0.007
Web extractor	per call when invoked	$0.007

Example request

curl https://api.empiriolabs.ai/v1/chat/completions \
  -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{"model": "fugu-max", "messages": [{"role":"user","content":"Hello"}]}'

Parameters

Parameter	Type	Required	Default	Description
`max_tokens`	integer	no	`32768`	Maximum number of output tokens for the final answer. Must be at least 16. The conductor needs room to work, so very small values can truncate or empty the answer. · Range: 16 – 131072
`reasoning_effort`	enum	no	`"high"`	How hard Fugu Max reasons. Reasoning is always on. The default is high; xhigh and max are aliases of the same maximum effort. · Allowed: `high`, `xhigh`, `max`
`tool_web_search`	boolean	no	false	Enable built-in web search. Adds $0.007 to the request cost for each invoked call, and a single request can run more than one call. Page fetches are billed at the same rate.
`tools`	array	no	`[]`	OpenAI-compatible function calling tool definitions.
`tool_choice`	object	no	-	OpenAI-compatible tool choice control.
`response_format`	enum	no	-	Return structured JSON output. JSON mode returns any valid JSON object; JSON Schema mode enforces the exact schema you provide.

Notes

Fugu Max is the cost-efficient multi-agent conductor: each request assembles a right-sized team of expert models for the task and composes their work into a single answer.

Latency and streaming

Simple prompts usually return in a few seconds, and heavier prompts take longer while the conductor works.
The full answer is returned all at once when the model finishes, not token by token. Streaming is accepted, but it delivers the complete response at the end rather than streaming tokens as they generate.
max_tokens must be at least 16. Leave generous headroom, since very small limits can truncate or empty the answer.

Capabilities

Text and image input, with a 1M token context.
Always-on reasoning. high is the default, xhigh raises the effort, and max is accepted as an alias of xhigh.
Function calling, strict JSON Schema structured output, and built-in web search that cites its sources when available.

Billing

Billed on full token usage, including the orchestration tokens the model uses internally, so even short prompts carry some cost.
One flat token rate at every context length, with no higher rate for long prompts.
Built-in web search and page fetch are billed per executed call at the rates shown, and a single request can run more than one call.

Per-tool billing (usage.tool_usage)

When this model invokes tools (web search, code interpreter, etc.) inside a single request, the response carries a normalized usage.tool_usage map alongside the token counts. The example below shows the shape, exact field names, units, and which tools appear can vary slightly per provider:

"usage": {
  "prompt_tokens": 123,
  "completion_tokens": 456,
  "cost_usd": 0.0042,
  "tool_usage": {"web_search": 3, "code_interpreter": 1}
}

The tool counts are already factored into cost_usd, they are surfaced for transparency so you can audit per-tool billing. The field is omitted when no tools were invoked.

Machine-readable schema: GET https://api.empiriolabs.ai/v1/models/fugu-max.