Edge0 released weights, an inference framework and deployment tools for its 145B and 120B models. The 145B model activates about 10.12B parameters per token and reports vendor evaluation scores of 64.6 on DeepSWE and about 86.5 on Terminal-Bench 2.1. The announcement claims 29GB peak laptop memory and over 80 tokens per second, while repository documentation separately lists GPU and host RAM usage and hardware-dependent speeds. Local execution can reduce cloud token fees, but total memory, storage, electricity and the limits of vendor-run evaluations still matter.