OpenMythos: 14.8k-star open reconstruction of the "Claude Mythos" architecture
tom_doerr · x · 2026-08-27
GitHub project OpenMythos by kyegomez, now at 14.8k stars, offers a theoretical reconstruction of Anthropic's rumored "Claude Mythos" architecture built from first principles using public research literature. It is not affiliated with Anthropic.
The implementation uses a Recurrent-Depth Transformer (RDT) with three stages: Prelude (transformer blocks), a looped Recurrent Block (up to maxloopiters), and a final Coda. Attention is switchable between MLA and GQA, and the feed-forward uses a sparse MoE with routed and shared experts — suited for exploring compute-adaptive, depth-variable reasoning. Installable via pip with optional Flash Attention 2 support.
Related event: OpenMythos: Open-Source Theoretical Recreation of Claude Architecture(3 posts)→
More from Research
- Interpretability paper incoming: in-depth look at testing explanations on modern LLMs — burny_tech · 2026-10-16
- Better-learning models tend to be more interpretable, argues Nick Cammarata — burny_tech · 2026-10-16
- 'Interp was mostly hopeless before — now it's automatable': Cammarata on interpretability — burny_tech · 2026-10-16
- Cornell and Amazon revisit the 16x pretraining compute gap: diffusion LMs need only 3-8x — LucaAmb · 2026-10-16
- AI trained on the fraction of human intelligence that fit the page — r0ck3t23 · 2026-10-16
- Quotient Recurrent Reasoners hit 97.9% on Sudoku-Extreme with 78% fewer loops — burny_tech · 2026-10-16