Anyone tried out the new “world-ending, Mythos-class” model?
Tried this out yesterday and my initial take is it feels like a significant cut above their previous flagship model (Opus). I also believe that everyone’s concerns of irresponsible or inefficient usage are only going to be exacerbated by this model…
For any Anthropic fans: The Fable model is available under a few of the subscription plans till June 22nd (after that they are moving it to a separately billed feature “until further notice”).
Models keep getting better, but the rate of improvement is increasing more slowly than prices, so most of us plebs are gonna be stuck at Opus or local model levels I think
Fable 5 @ Medium - more bang for your buck token wise vs. Opus 4.8 @ xhigh
In my usage so far, Fable 5 steers itself better than Opus 4.8 - but if you do a separate dedicated planning pass in 4.8 before an Act pass in 4.7 or 4.6, you can achieve the same output, less token usage.
My workflow is generally to plan in Claude 4.8 XHigh, then execute in a mixture of Claude, Codex, Local LLM. The executors don’t need so much “thinking” when the planning is pre-baked.
