Aethera trains language models that prove what they claim — every answer carries a formal derivation you can check. Pass the light across the marble to see the reasoning lattice underneath.
92.1% miniF2F, pass@1
68.4% ProofNet
1.3B weights, public