What is Instruction tuning?
Also called: chat model · aligned model · RLHF
Instruction tuning is the training phase that turns a raw text predictor into an assistant that follows requests, stops appropriately and behaves — the difference between a base LLM and a chat model.
After pre-training, a base model just continues text. Vendors then train it on examples of instructions and good responses, and on human feedback about which answers are better (RLHF). The result treats your message as a request, answers it, and adopts the familiar assistant voice.
Because this layer is thin relative to pre-training, you can steer through it: ask for a different voice, forbid the preamble, demand a format. Different vendors tune differently, which is why the same prompt behaves slightly differently on Claude, GPT and Gemini.
- Override the default voice explicitly when you don't want it.
- Learn one structure; adapt the surface per model.
Use it right now
Ask our brain anything on the homepage — it remembers the whole conversation — or write a brief in the Studio and see the prompt it compiles to.