This is useful data. Our current harness uses Opus 4.8 x2-4 (scale depending on task). We had 72 hours of glory with Fable x2-4, but that’s gone now, and Opus isn’t cutting it.
Biggest barrier is modifying the harness for OpenAI generic vs Anthropic. Also… I’m probably out of touch here, but can other models reliably use tools now? That was a big reason for settling with Anthropic.
Introducing the Fusion API, the smartest compound model in the market.
Fusion achieves Fable-level intelligence at half the price.
How it works 👇