Sakana's Fugu Ultra orchestrator lands on Vercel AI Gateway
- Source
- vercel.com
- Author
- Sakana AI
- Date

Context
Developers who use Vercel's AI Gateway can now call Sakana AI's Fugu Ultra by setting the model name to sakana/fugu-ultra, exactly as they would pick any single model there. What answers the request is not one model. According to Vercel's changelog, Fugu Ultra is built on a pool of publicly accessible frontier models: it sends each problem to between one and three agents, depending on the problem, and combines their results into a single answer. Vercel says that on reasoning and scientific benchmarks its capabilities are similar to those of Anthropic's Claude Mythos Preview and Fable 5, without naming the benchmarks or giving scores.
The developer chooses Fugu Ultra but not the models behind it. Sakana's Fugu page says the system grows out of its research on learned orchestration, including a coordinating model trained with reinforcement learning to discover how to divide work among other models, and that the specific models Fugu selects and how it coordinates them are proprietary. Sakana's own pitch for Fugu as a multi-agent system behind one model API covers that design in more detail.
The Gateway listing changes how easily that system can be tried. AI Gateway already gives developers one API for many providers, with usage and cost tracking, retries and failover, and Vercel says it passes provider pricing through with no markup or platform fee. A team already on the Gateway can therefore compare Fugu Ultra with the single models it uses now by changing one string, and see its usage and cost in the same reports.
- multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
- inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
Checking sign-in…
Loading comments…





