Sakana AI Launches Fugu Multi-Agent System with Superior Benchmark Performance

KuCoinFlash

Release Time: 06/22/2026 07:09:12

Summary

Sakana AI, a Japanese company, has launched its commercial product Sakana Fugu and opened early beta access. The multi-agent system supports OpenAI-style APIs and comes in two versions: Mini for low-latency applications and Ultra for complex tasks. Built on ICLR 2026 research, it employs a self-learning model for dynamic task allocation. On-chain reports indicate that Sakana Fugu Ultra outperformed GPT 5.4, Gemini 3.1, and Opus 4.6 on challenging reasoning and coding benchmarks such as GPQAD and SWEPro. AI and crypto news highlights the company’s focus on high-performance AI solutions.

ME AI News, according to monitoring by Beating, Japanese AI research company Sakana AI has announced the launch of its flagship commercial product, Sakana Fugu, and opened early Beta test applications. The system is positioned as a multi-agent orchestration platform, offering services via a single-model API compatible with OpenAI’s format, including Sakana Fugu Mini, optimized for low latency, and Sakana Fugu Ultra, designed for demanding tasks. The architecture is based on the team’s Trinity and Conductor papers published at ICLR 2026. At its core is a lightweight, self-learning language model that does not rely on pre-defined team roles or fixed pipelines, but dynamically invokes and assigns tasks from a pool of Worker models based on task complexity. The system also supports test-time scaling and features adaptive recursive invocation capabilities. The model can read its own prior outputs as context, enabling it to autonomously identify flaws during runtime and initiate correction workflows; users can adjust recursive depth as a configurable compute axis during inference. Evaluation data shows that Sakana Fugu Ultra outperforms leading single models in rigorous reasoning and coding benchmarks. In GPQAD, LCBv6, and SWEPro tests, Sakana Fugu Ultra achieved scores of 95.1, 93.2, and 54.2 respectively, surpassing mainstream flagship single models such as GPT-5.4, Gemini 3.1, and Opus 4.6. (Source: BlockBeats)

Source:Show original

Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.