Sakana AI: Rolls Out Fugu Multi-Agent LLM Orchestrator
Sakana AI Fugu Ultra hits 73.7 on SWE-Bench Pro, matching Fable 5 via recursive multi-agent LLM orchestration for coding and reasoning tasks.
SourceAnalysis
Sakana AI released Fugu, a single-model API that orchestrates pools of LLMs including recursive self-calls, assigning Thinker, Worker and Verifier roles across coding, math and reasoning workloads. Fugu Ultra posted 73.7 on SWE-Bench Pro, level with Fable 5 and Mythos, while one engineer logged over twenty code-review issues versus the usual three from prior tools. The architecture delivers frontier multi-agent LLM orchestration without export-control exposure.
Miles Deutscher
@milesdeutscherCrypto analyst. Busy finding the next 100x.