Sakana AI has released Fugu Ultra v1.1, an updated version of its AI model router that the company claims outperforms Anthropic's Claude Fable 5 without even routing requests to it. The update reportedly delivers performance gains of up to 7.9 points over the previous version.
Fugu Ultra functions as a routing layer that directs queries to different AI models based on task complexity and requirements. Rather than running a single model, it intelligently selects from a pool of available models to optimize speed and cost. Sakana's claim centers on the router's ability to exceed Fable 5's performance using its existing model pool, suggesting the routing algorithm itself has improved significantly.
The v1.1 update introduces a Claude Code-compatible endpoint, enabling developers to swap Fugu Ultra into existing workflows built for Anthropic's API without major integration changes. This compatibility move targets developers already invested in Claude ecosystems.
However, the performance claims lack independent verification at this stage. Sakana hasn't published detailed benchmarks showing test conditions, datasets, or model combinations used in the comparison. The claim that Fugu Ultra beats Fable 5 without including it in the routing pool raises questions about which models were actually used and under what circumstances the router achieved those results.
The service remains unavailable in the European Union, likely due to regulatory requirements under the EU AI Act and data protection frameworks. This geographic limitation constrains Sakana's addressable market and developer access in one of the world's largest tech regions.
Sakana's approach reflects a broader industry trend where routing and mixture-of-experts layers become competitive products themselves. If verified, this update would demonstrate that intelligent routing can outperform single model responses. The Claude Code compatibility signals Sakana's strategy to position Fugu Ultra as a drop-in replacement rather than a fundamental architecture change