Pruna AI builds a model-optimization framework that applies efficiency methods such as caching, pruning, quantization and distillation to shrink machine-learning models and speed up inference while measuring any resulting quality loss. The company positions efficiency as a sustainability play, reporting large reductions in the compute and energy needed to run models. In March 2025 Pruna open-sourced its framework, having demonstrated making a Llama model roughly eight times smaller, and it also sells an enterprise edition with advanced features including an optimization agent. It was founded in 2023 by Bertrand Charpentier and John Rachwan.
Funding
Pruna AI raised a $6.5 million (about EUR 6.2 million) seed round from investors including EQT Ventures, Daphni, Motier Ventures and Kima Ventures.