Fugu Ultra v2 is the higher-performance model in Sakana AI's Fugu family and the direct successor to Fugu Ultra. Rather than a standard monolithic language model, it is 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 is optimised for complex, multi-step reasoning, coding and agentic workflows. Orchestration tokens are billed as standard input and output tokens. Compared with Fugu Ultra it adds document (file) input alongside text and images, and reasoning is always on.
Coordinates expert agents across a large codebase, surfacing issues that a single-pass review misses.
From a single instruction, reads a paper, implements it, runs training and analyses the gap between reported and reproduced results.
The 1,000,000-token context window plus file input suits contract review, due diligence and multi-document synthesis in one request.
Always-on reasoning, tool calling and web search make it suited to long-running autonomous tasks rather than single-turn chat.
Requests above 272,000 tokens are billed at a higher tier: $13.754 / 1M input, $1.3754 / 1M cached input and $61.893 / 1M output.
Released September 11, 2026.
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
Complex multi-step reasoning is a documented use case for Fugu Ultra v2. It is also documented as suited to autonomous research and full-stack software development.
It is documented for coding-related work, including real-world software engineering and code review. It is also documented for visual reasoning and data interpretation, including chart analysis and document or chart analysis.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "sakana/fugu-ultra-v2", "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-ultra-v2", "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-ultra-v2","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Cached input |
Requests above 272,000 tokens are billed at a higher tier: $13.754 / 1M input, $1.3754 / 1M cached input, $61.893 / 1M output.
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Fugu Ultra v2 This page | Deep reasoning + long-context agents | |||
| Reasoning + agents | ||||
| Reasoning + agents | ||||
| Balanced coding + agents | ||||
| Reasoning + agents |
Fugu Ultra v2 is the higher-performance model in Sakana AI's Fugu family and the successor to Fugu Ultra. Rather than a monolithic language model, it is a learned multi-agent orchestration system that routes tasks across a swappable pool of underlying models and recursively calls instances of itself.
It adds document (file) input alongside text and images, and reasoning is always on. The 1,000,000-token context window and the orchestration approach carry over unchanged.
Sakana AI.
1,000,000 tokens, with up to 128,000 tokens of output.
Orchestration tokens are billed as standard input and output tokens. Requests above 272,000 tokens move to a higher pricing tier.
It takes image, text and file input and returns text.
Yes. It supports function calling, structured outputs, streaming and web search.
Complex, multi-step work where depth matters more than latency: thorough code review, research and paper reproduction, long-document synthesis, and long-running agentic workflows.
Send a request to the chat completions endpoint with the model id sakana/fugu-ultra-v2. The model also answers to the alias fugu-ultra-v2.
