How to Fine-Tune an Open-Weights LLM: A Hands-On Guide Using Inkling

tldr: Fine-tuning an LLM means continuing to train an existing model on your own data so it specializes in your task. With an open-weights model like Thinking Machines’ Inkling (975B parameters, 41B active), you can do this with LoRA (on hosted fine-tuning infrastructure for the full model, or a single multi-GPU node for its smaller sibling) instead of retraining from scratch. But fine-tune only after prompt engineering and RAG have failed you; it’s the most expensive and least flexible of the three, and most teams reach for it too early.