Product · mentioned 63 times
AI model developed by Anthropic, a frontier large language model used for coding, design and enterprise tasks.
Tap a name to go there · tap a line to see why they are linked
Latent Space: The AI Engineer Podcast· 7 Nov
swyx illustrates with a hypothetical that Anthropic could kill Harvey by only using Claude Fable 6 for legal work.MTS· 7 Nov
Herbie Bradley says that when agent traces are published for models such as Astra or Fable, examining them closely often shows the model has simply hacked the benchmark in the published trace.etn.· 6 Nov
Anish says Anthropic distils Mythos into smaller models like Opus and Fable, which it releases, so it does not have to redo its pre-training runs.MTS· 6 Nov
Qasim Wani says models including Astra and Fable were given the GDPX reward-hacking detection task.The Frontier with Brendan Foody· 5 Nov
Andy Fang says some people at the labs might think Fable, Mythos or Soul is good enough to file your taxes, and that the only blocker might be access to the data.