Sakana AI launches Fugu, a multiagent orchestration system accessible via an OpenAI-compatible API, claiming Fugu Ultra matches Fable and Mythos on benchmarks
Last night, the increasingly enterprise-focused AI startup Sakana launched Fugu, a multi-agent orchestration system …
VentureBeatCarl Franzen
Context & Ripple Effects
Sakana AI previously built Japanese-language models through model merging, then broadened from corporate offerings with its Sakana Chat launch as localization competition intensified. Fugu returns the company’s focus to an enterprise-oriented capability: packaging multi-agent work behind a single API.
Related coverage places Fugu alongside claims from other Asian AI developers to approach Anthropic models that are constrained by US export restrictions. That makes the launch relevant not only as a model-performance claim, but as an alternative access path for customers with limited access to leading foreign systems.
First-order effects
Developers can access Sakana’s multi-agent orchestration through an OpenAI-compatible API, lowering integration friction for teams already built around that interface.
Sakana is positioning Fugu Ultra directly against Fable and Mythos on benchmarks, raising the immediate burden on the company to substantiate performance in customer deployments.
Second-order effects
Enterprise buyers evaluating agentic workflows gain another vendor option, particularly where access to leading US models is restricted or localization is a priority.
Competing model providers and orchestration vendors face pressure to offer comparable API compatibility and to distinguish themselves on reliability, task coordination, and deployment support rather than raw benchmark claims alone.
Third-order effects
If multi-agent systems are increasingly sold through familiar single-model APIs, orchestration may become a product layer that customers can swap independently of the underlying models.
Export limits and local-market competition could accelerate regional AI stacks, though benchmark parity claims will matter structurally only if they translate into sustained real-world performance and adoption.
The trend: This is one data point in the shift from standalone foundation models toward API-delivered agent orchestration, with regional providers seeking to fill gaps created by localization demands and uneven model access.
Another new idea to push the state of AI architectures forward. Sakana released a model that effectively uses a mixture of models to get work done. You get a single API but then the work gets farmed out the model that best performs the task. “Fugu manages model selection, [image]
Human intelligence is fundamentally a collective intelligence. We solve complex problems by participating in a vast cultural network that builds upon ideas across generations. I believe the strongest AI systems will become a collective intelligence, too. Since we started Sakana
Sakana Fugu Ultra is live on AI Gateway. Mythos-class intelligence in a single call, with a whole pool of models behind it. 𝚖𝚘𝚍𝚎𝚕: ‘𝚜𝚊𝚔𝚊𝚗𝚊/𝚏𝚞𝚐𝚞-𝚞𝚕𝚝𝚛𝚊’ https://vercel.com/...
Major kudos to @hardmaru and @SakanaAILabs on the launch 🚀 “our core conviction has been that the most powerful AI systems will be collaborative ecosystems, not isolated monoliths...the future belongs to systems that explicitly learn how to coordinate collective intelligence.”
I have been trying Sakana Fugu Ultra-high and, first, it is incredibly slow: my typical coding tests (shaders, interactive scenes) take 30 minutes to run And the results are... fine. It does not match Fable in real use. Its harbor is a good example: https://ai-harbor-town-gallery…
I really do not believe that ‘Fugu Ultra’ “matches the performance of Fable and Mythos” in any meaningful way. (1) @SakanaAILabs has a past track record of hype that ends up being nonsense and (2) I just don't think they have a plausible pathway to this level of capability.
to be clear, this is a closed source orchestrator on top of closed source models. if before you didn't control the models, now you don't even control which ones are used or how much. this is not “AI sovereignty” i've also read the tech report to get an opinion on the technical
Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API. Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability without the risk of export controls. Try it: https://sakana.ai/fugu/ 🐡 [image]
Tried & liked it on https://archive.tw/. Fugu Ultra pairs well as a advisor & planner with Composer 2.5. For scope/architecture, it's on par with Fable orchestration. Advisor doesn't slow the loop if the driver stays fast & https://omp.sh/ can split it from worker.
SAKANA FUGU ULTRA vs. CLAUDE OPUS 4.8 RESULTS Prompt: “build a really high quality single html file crossy road game with three.js” Sakana Fugu Ultra: - Tokens Used: ~89k ($7.32) - Time Elapsed: 22 minutes - Issues: inverted directional turn, wonky camera, no sfx, not [video]
Sakana's Multi-Agent on par with Fable 5: Sakana AI's Fugu Ultra may not be a new frontier model in the classical sense. It is more like a learned orchestration layer that turns multiple frontier models into a coordinated agent team. The next jump in AI capability may come [image…
> without the risk of export controls lol this is a closed source llm orchestrator that relies partly on closed source models api, they don't even report what percentage of closed vs open models the system uses to achieve these scores on benchmarks, this is very misleading
Sakana Fugu — a multiagent system for general tasks that performs on par with the Mythos Preview — It uses a collection of open and closed models, including itself, to handle all angles of a problem — model selection, delegation, verification and synthesis — sakana.ai/fugu-re…