CompactifAI
SpainModel developer (compression)
Quantum-inspired compression making big models radically smaller.
multiversecomputing.com/compactifai
What they do
CompactifAI is Multiverse Computing's flagship product, using tensor-network compression to shrink LLMs by up to ~93% in memory with only small accuracy loss; the compressed models (and HyperNova line) can run on modest hardware. Based in San Sebastián; raised a €189M Series B.
Models they develop
1 in this directory.