Prime Lab expands beyond LoRA: full fine-tuning, pretraining and dedicated compute tiers

willcb · x · 2026-11-19

xeophon shared a Prime Lab vs Model Factory feature comparison: LoRA RL supported across all tiers, while full fine-tuning, SFT, full-parameter RL, and pre/mid-training are exclusive to Model Factory's on-demand and dedicated tiers. Dedicated tier adds custom framework forks and self-healing compute with Slurm/Kubernetes access. The pitch: own your intelligence with a full-stack training platform beyond the original LoRA-only Lab.

Original post →

More from Infra

Infra channel →