Sakana Fugu Ultra Is Now Available in Puter.js

Puter.js now supports Fugu Ultra, the multi-agent orchestration model from Tokyo-based Sakana AI — and you can use it without any API keys.

What is Fugu Ultra?

Fugu Ultra is a learned orchestration system that routes each task across a swappable pool of frontier LLMs behind a single OpenAI-compatible endpoint. Rather than relying on one monolithic model, it assigns Thinker, Worker, and Verifier roles to specialist agents, then synthesizes their outputs into a single response. The coordination is grounded in Sakana's TRINITY and Conductor research, both published at ICLR 2026. Key highlights include:

• Multi-Agent Orchestration: A conductor layer classifies each query and decides how many expert agents to call, coordinating them over several turns before producing a final answer
• Frontier Benchmark Results: Scores 93.2 on LiveCodeBench, 90.8 on LiveCodeBench Pro, and 95.5 on GPQA-Diamond, with 73.7 on SWE-Bench Pro and 50.0 on Humanity's Last Exam
• 1M Token Context Window: Process entire codebases, long documents, or extended agent traces in a single request, with up to 128K output tokens
• No Vendor Lock-In: The agent pool is swappable, so you get frontier-level quality without committing to a single model provider
• Built for Hard Tasks: Early users reach for it on complex reasoning, code review, cybersecurity analysis, paper reproduction, and patent and literature research

Because Fugu Ultra coordinates multiple models per query, it trades response time for depth. Reach for it when accuracy on a demanding, multi-step problem matters more than latency.

Examples

Complex multi-step reasoning

const response = await puter.ai.chat(
  "A distributed cache shows 30% higher p99 latency after adding a third replica. Walk through the likely causes and how you would isolate each one.",
  { model: 'sakana/fugu-ultra', stream: true }
);

for await (const part of response) {
  puter.print(part?.text);
}

Code review

puter.ai.chat(`Review this function for correctness and edge cases:

function chunk(arr, size) {
  const out = [];
  for (let i = 0; i < arr.length; i += size) {
    out.push(arr.slice(i, i + size));
  }
  return out;
}`,
  { model: 'sakana/fugu-ultra' }
);

Cybersecurity analysis

puter.ai.chat("Analyze this Express route for security issues and suggest fixes: app.get('/files', (req, res) => res.sendFile(req.query.path))",
  { model: 'sakana/fugu-ultra' }
);

Research and reproduction

puter.ai.chat("Outline a plan to reproduce the main result of a paper that proposes a new learning-rate schedule, including the experiments and ablations you would run to validate it.",
  { model: 'sakana/fugu-ultra', stream: true }
);

Get Started Now

Just add one library to your project:

// npm install @heyputer/puter.js
import { puter } from '@heyputer/puter.js';

Or add one script tag to your HTML:

<script src="https://js.puter.com/v2/"></script>

No API keys needed. Start building with Fugu Ultra immediately.

Learn more:

• Free, Unlimited Sakana AI API
• Puter.js AI Documentation