Sakana: Fugu Max
sakana/fugu-max · SakanaUnmeasurable- Declared family
- Unknown
- Entered catalogue
- 11 September 2026
- Last tested
- 5 October 2026
- Measurement
- Unmeasurable: prompt caching corrupts its token accounting
- 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 | A combination of settings no other listing in the catalogue uses | api_0d2351af |
| 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, high by default | its own listing |
| Who served our requests | Sakana AI | we measured it, 5 October 2026 |
Closest measured models
Agreement across the mutually clean test stringsGoogle: Lyria 3 Clip Previewgoogle/lyria-3-clip-preview1/25 · 4%Ox Alphastealth/ox-alpha0/25 · 0%Z.ai: GLM 5.3z-ai/glm-5.30/25 · 0%DeepSeek: DeepSeek V4 Pro 0813deepseek/deepseek-v4-pro-08130/25 · 0%Xiaomi: MiMo-V2.5xiaomi/mimo-v2.50/25 · 0%Xiaomi: MiMo-V2.5-Proxiaomi/mimo-v2.5-pro0/25 · 0%
Click a row to open the full comparison.
The measured fingerprint
What each test string cost this model, in tokensWe sent Sakana: Fugu Max 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, 25 excluded as corrupt
English and whitespace
en-prose53
en-long62
spaces-209
spaces-60x
tabs-2010
newlines-2012
mixed-wsx
Digits
digits-917
digits-12x
digits-30x
digits-sepx
float-longx
CJK
zh-common37
zh-long46
zh-rarex
ja-kana33
ja-kanji44
ko49
Other scripts
ru46
ar41
hex
hi61
th53
el54
Emoji
emoji-basic25
emoji-skin44
emoji-zwj-family33
emoji-zwj-x3x
emoji-flagsx
emoji-profx
Rare Unicode
mathx
boxdrawx
combiningx
cjk-ext-bx
surrogatesx
zalgox
rtl-mix21
Code
py-codex
py-indent60
jsonx
htmlx
regexx
camel21
snake28
Repetition and encodings
rare-word-x5111
repeat-tok104
base64x
hexx
urlx
uuidx
An x marks a row excluded as corrupt (caching or provider interference detected during measurement). Overhead subtracted: 429 prompt tokens.
Declared record
What the catalogue claims about this modelFugu Max is the cost-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+file->text
- Declared tokenizer
- Other
- Prompt price
- $2.00 / M tokens
- Completion price
- $6.00 / M tokens
Supported parameters
include_reasoningreasoningreasoning_effortstructured_outputstool_choicetoolsweb_search_options
defaults: {"temperature":null,"top_p":null,"top_k":null,"frequency_penalty":null,"presence_penalty":null,"repetition_penalty":null}
History
Every observation, kept as taken- 11 September 2026Enters the OpenRouter catalogue with no declared family.
- 5 October 2026Fingerprinted in the YFarmX catalogue sweep · 25 of 50 strings measured clean.
CompareCompare now