Sakana: Fugu Ultra

sakana/fugu-ultra · SakanaNot yet established

Declared family: Unknown
Entered catalogue: 24 June 2026
Last tested: 17 October 2026
Measurement: Deferred: spend cap reached mid-measurement
Evidence confidence: Unknown

What we checked

Four different kinds of evidence, and what each one showed

We checked	What we found	Where it came from
How it counts tokens	Not measured yet	we measured it
How its API is set up	Set up identically to 1 other listing, field for field	api_956e60ef
How much it can read and write	Reads up to 1,000,000 tokens at once, writes up to 128,000	its own listing
Whether it thinks before answering	Always reasons, and cannot be turned off, at max, xhigh, high, xhigh by default	its own listing
Who served our requests	Sakana AI	we measured it, 17 October 2026

The measured fingerprint

What each test string cost this model, in tokens

We sent Sakana: Fugu Ultra fifty short pieces of text and recorded what each one cost it in tokens. The bars below are those costs. Two models built on the same tokenizer produce the same bars; a model built on a different one produces a different set, which is what makes this a fingerprint.

All 50 rows, 35 excluded as corrupt

English and whitespace

en-prose14

en-longx

spaces-203

spaces-603

tabs-203

newlines-204

mixed-ws6

Digits

digits-94

digits-125

digits-3011

digits-sep8

float-long10

CJK

zh-common13

zh-longx

zh-rarex

ja-kana11

ja-kanji13

ko11

Other scripts

rux

arx

Emoji

emoji-basic

emoji-skin

emoji-zwj-family

emoji-zwj-x3

emoji-flags

emoji-prof

Rare Unicode

math

boxdraw

combining

cjk-ext-b

surrogates

zalgo

rtl-mix

Code

py-code

py-indent

json

html

regex

camel

snake

Repetition and encodings

rare-word-x5

repeat-tok

base64

hex

url

uuid

An x marks a row excluded as corrupt (caching or provider interference detected during measurement). Overhead subtracted: 1332 prompt tokens.

Declared record

What the catalogue claims about this model

Fugu Ultra is the higher-performance model in Sakana AI's Fugu family. Rather than a single monolithic model, Fugu is a learned multi-agent orchestration system: a language model trained to route...

Modalities: text+image->text
Declared tokenizer: Other
Prompt price: $5.00 / M tokens
Completion price: $30.0 / M tokens

Supported parameters

include_reasoningreasoningreasoning_effortstructured_outputstool_choicetoolsweb_search_options

Default settings

Temperature: Unspecified
Top p: Unspecified
Top k: Unspecified
Frequency penalty: Unspecified
Presence penalty: Unspecified
Repetition penalty: Unspecified

Catalogue entry

History

Every observation, kept as taken

24 June 2026Enters the OpenRouter catalogue with no declared family.
17 October 2026Fingerprinted in the YFarmX catalogue sweep · 15 of 50 strings measured clean.