Sakana AIText Generation/v1/ai/chatReasoning

Fugu Ultra v2 API

Access Sakana AI's Fugu Ultra v2 through one API key. The top tier of the Fugu multi-agent system.

Sakana AI's higher-performance Fugu model. A learned multi-agent orchestration system that routes work across stronger specialist models, with a 1M token context window and 128K output tokens.

Full Docs Get API Key

Takes 30 seconds.

Fugu Ultra v2 API pricing

Pay per request from your YepAPI balance — no subscription, no monthly minimum. Prices updated September 27, 2026.

Token type	Price
Input	$7.38 / 1M tokens
Output (incl. reasoning)	$44.31 / 1M tokens
Context window	1M tokens

What it costs	Per job	Per 1,000
Chat reply1K input + 500 output tokens	$0.03	$29.54
RAG answer8K input + 800 output tokens	$0.094	$94.49
Agent step with tools30K input + 2K output tokens	$0.31	$310.02
Long document summary100K input + 2K output tokens	$0.827	$826.62

4,900+

Developers

2.3M+

API calls served

184

Endpoints

$0.01

Per call

Yep, that's it.

Try it live

Send a message and see Fugu Ultra v2 respond in real time.

POST/v1/ai/chatsakana/fugu-ultra-v2

Message *

Max Tokens

Maximum tokens in the response.

Stream

Real-time tokens

Hit "Send Request" to see the response

Fugu Ultra v2 specs

Provider	Sakana AI
Model ID	sakana/fugu-ultra-v2
Endpoint	/v1/ai/chat
Type	Text Generation
Context window	1M tokens
Max output	128K tokens
Reasoning	Yes
OpenAI-compatible	Yes — /v1/ai/chat/completions

How to get a Fugu Ultra v2 API key

1
Create an account
Sign up at yepapi.com with Google or email. It takes about 30 seconds.
2
Create an API key
Open the dashboard, go to API Keys and create a key. Copy it — it starts with yep_sk_.
3
Call Fugu Ultra v2
Send a request to /v1/ai/chat with the x-api-key header and model "sakana/fugu-ultra-v2".
4
Pay as you go
Usage is billed per request from your balance. Top up any amount; there is no subscription.

Strengths

✓

Top of the Fugu family

Fugu Ultra v2 is the higher-performance tier of Sakana AI's Fugu system, orchestrating stronger specialist agents than Fugu Max for the hardest reasoning and engineering tasks.

✓

Multi-agent orchestration

A router model decomposes each task, dispatches parts to specialist agents, and merges the results — an architecture aimed at problems that benefit from several perspectives.

✓

1M token context, 128K output

Handles up to 1,000,000 input tokens and returns up to 128,000 output tokens per call, so whole codebases and long agent histories fit in one request.

✓

Tools, web search and structured output

Supports function calling, JSON schema structured outputs, reasoning effort control, and built-in web search options.

Quick start

Copy this snippet and start making calls with Fugu Ultra v2.

const res = await fetch('https://api.yepapi.com/v1/ai/chat', {
  method: 'POST',
  headers: {
    'x-api-key': 'YOUR_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    "model": "sakana/fugu-ultra-v2",
    "messages": [
      {
        "role": "user",
        "content": "Explain API gateways in 2 sentences."
      }
    ],
    "maxTokens": 256
  }),
});
const { data } = await res.json();
console.log(data.message.content);

Fugu Ultra v2 vs similar models

Every model below runs on the same API key — switch by changing the model ID.

Model	Provider	Price	Context
Fugu Ultra v2	Sakana AI	$7.38 in / $44.31 out per 1M	1M
Fugu Max	Sakana AI	$2.95 in / $8.86 out per 1M	1M
Nemotron 3 Super	NVIDIA	$0 in / $0 out per 1M	262K
Qwen3.7 Flash	Qwen	$0.04 in / $0.19 out per 1M	1M
Solar Mini 4	Upstage	$0.07 in / $0.30 out per 1M	524K
Nemotron 3.5 Lightning	NVIDIA	$0.10 in / $0.30 out per 1M	262K
DeepSeek V4 Flash	DeepSeek	$0.13 in / $0.27 out per 1M	1.0M
Qwen 3.5 Flash	Qwen	$0.10 in / $0.38 out per 1M	1M

Why use Fugu Ultra v2 through YepAPI?

✓One API key for all models — no separate accounts

✓OpenAI SDK compatible — just change the base URL

✓No monthly minimums — pay per token

✓Switch models with one line of code

✓Full provider passthrough — citations, search results, and all extras included

✓Streaming and non-streaming support on every model

✓Works with Cursor, Claude, LangChain, and any LLM tool

✓Unified billing across all providers

Fugu Ultra v2 API — pricing, context window & access

Fugu Ultra v2 is the higher-performance model in Sakana AI's Fugu family and the current revision of Fugu Ultra. Fugu is a learned multi-agent orchestration system: a language model trained to route parts of a task to specialist agents and combine their answers. Ultra v2 pairs stronger agents with a 1,000,000-token context window and up to 128,000 output tokens per response.

Through YepAPI you call Fugu Ultra v2 on an OpenAI-compatible endpoint with a single key — the same key that reaches GPT-6, Claude, Gemini and the SEO, SERP and scraping APIs.

What is Fugu Ultra v2?

Fugu Ultra v2 is Sakana AI's flagship Fugu tier. Where Fugu Max optimises for cost, Ultra v2 orchestrates the strongest specialist models in the system, so multi-step reasoning, code generation and research synthesis come back with more depth per call. It supersedes the original Fugu Ultra, which remains available upstream. The 128K output ceiling lets it return complete implementations or long reports in one response.

Build with Fugu Ultra v2 via YepAPI

YepAPI serves Fugu Ultra v2 on the OpenAI-compatible /v1/ai/chat endpoint. Point your base URL at YepAPI, add your key, and set the model string to fugu-ultra (or sakana/fugu-ultra-v2). Function calling, structured outputs and reasoning effort pass through unchanged. Dropping to Fugu Max for routine steps is a one-string change per request.

Fugu Ultra v2 API pricing — $7.38 / 1M input, $44.31 / 1M output

Fugu Ultra v2 costs $7.38 per 1M input tokens and $44.31 per 1M output tokens through YepAPI, pay per token with no minimums. That is the same price band as GPT-5.5, so treat it as a premium model: reserve it for tasks where the multi-agent depth pays for itself and route lighter work to Fugu Max.

Fugu Ultra v2 for hard reasoning and engineering

Ultra v2 suits problems that benefit from several angles of attack: architecture decisions, difficult debugging across a large codebase, research questions that need sources reconciled, and analyses where the reasoning must be shown. Built-in web search options ground its answers, and the 1M-token window keeps everything an agent has seen in context.

Start generating in 30 seconds

Pay per call, no subscriptions.

Get API Key

What developers say

“Switched from SerpAPI and cut our SERP costs by 80%. Same data quality, way simpler billing.”
Marcus T.
SEO Platform Founder

“One API key for AI models, SERP data, and web scraping. Saved us from managing 4 separate providers.”
Priya S.
Full-Stack Developer

“The $5 free credit let us prototype our entire rank tracking feature before committing. No other API does that.”
Jake R.
Indie Hacker

Frequently asked questions

Sakana AI's higher-performance Fugu model. A learned multi-agent orchestration system that routes work across stronger specialist models, with a 1M token context window and 128K output tokens.

Fugu Ultra v2 costs $7.38 per 1M input tokens and $44.31 per 1M output tokens through YepAPI, billed per request with no monthly fee. For example, a chat reply (1K input + 500 output tokens) costs about $0.03.

Sign up at yepapi.com (Google or email, about 30 seconds), open the dashboard and create an API key. The same key works for Fugu Ultra v2 and every other model — no separate Sakana AI account or waitlist.

Use "sakana/fugu-ultra-v2" as the model on /v1/ai/chat. The short alias "fugu-ultra" also works.

Yes. Point the OpenAI SDK at https://api.yepapi.com/v1/ai with your YepAPI key and set model to "sakana/fugu-ultra-v2". Streaming works the same way, so existing OpenAI code only needs the base URL and key changed.

Fugu Ultra v2 supports a 1M token context window with up to 128K output tokens per request.

Ready to use Fugu Ultra v2?

Pay per call, no subscriptions.

Get API Key Read Docs

Explore more models

Fugu Max

Sakana AI

Access Sakana AI's Fugu Max through one API key. Multi-agent orchestration at a workhorse price.

1M ctx$2.95/M in

Seed 2.0 Lite

ByteDance

Access Seed 2.0 Lite through one API key. ByteDance's latest bilingual model.

262K ctx$0.37/M in

Claude Haiku 5.5

Anthropic

Access Claude Haiku 5.5 through one API key. Anthropic Fast Low-Cost Claude.

1M ctx$0.15/M in

Kimi K2.7 Code

Moonshot AI

Access Kimi K2.7 Code through one API key. Moonshot AI's coding-specialized model.

262K ctx$0.99/M in

Gemini 3.5 Flash Lite

Google

Access Gemini 3.5 Flash Lite through one API key. Google Cheapest Gemini Subagent Model.

1.0M ctx$0.44/M in

MiniMax M3.1 Flash

MiniMax

Access MiniMax M3.1 Flash through one API key. MiniMax Fast Multimodal Agent Model.

1.0M ctx$0.22/M in

More Text Generation models

All Sakana AI models →All AI models →