Anthropic's Mythos-class model allegedly leaked on Hugging Face ๐
retained text
A Hugging Face repo called `burner22/numbat` claims to contain an internal Anthropic Mythos-class training checkpoint. Claimed specs: * 14 trillion parameters * 1,224 MoE experts * 7.72 TB of weights * 772 weight shards * Hybrid GDN + MLA architecture The repo also lists model architecture code, configs, and training metadata. No independent verification yet that these are genuine Anthropic weights. If this turns out to be real, it's massive. ๐ Anyone looked into the actual files yet? Legit leak or elaborate fake? [https://huggingface.co/burner22/numbat](https://huggingface.co/burner22/numbat)
document fetched 2026-11-02 00:22 ยท canonical https://reddit.com/r/ClaudeAI/comments/1xl47f5/anthropics_mythosclass_model_allegedly_leaked_on