🔥Anthropic Mythos benchmark score is four times that of GPT-5.4, with community speculation that it uses a byte architecture


According to 1M AI News monitoring, Anthropic refuses to disclose the model architecture of Mythos, but a test result has sparked speculation. The official system card shows that Mythos scored 80.0% in the GraphWalks BFS test, Opus 4.6 scored 38.7%, and GPT-5.4 only 21.4%. The differences on other benchmarks are smaller. Meta machine learning engineer Chris Hayduk pointed out that the unusual spike indicates a cyclic language model architecture. ByteDance's Seed team published the LoopLM paper last October, with the core…
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pin