Posted inAI News
Sakana Fugu Tops Benchmarks but Runs Slow
Sakana Fugu orchestrates several models and matches Fable 5 on benchmarks, but Ethan Mollick calls it very slow, around thirty minutes per coding test.
Articles tagged Sakana AI
Posted inHorizon Labs Our AI Tool Tests
Sakana AI Marlin Test: What the 8-Hour Agent Delivers
Marlin, Sakana AI's agent, runs for 8 hours straight and returns a full strategy. We put it to the test: real capabilities, limits, and verdict.