Claude Quiz: From Haiku to Opus Models

Claude was developed by a company founded in 2021 by researchers who had previously worked at a prominent AI laboratory. The organisation was built around the belief that as AI grows more capable, safety research must keep pace. It has raised significant funding while maintaining an independent research focus.

Claude's developers created a method that does not rely solely on human feedback for every decision. Instead, the model is guided by a written set of principles during training, allowing it to evaluate its own responses. This approach reduces dependence on large volumes of human-labelled data for safety.

Claude's model family uses a naming system drawn from the Greek alphabet to distinguish capability levels. The highest tier carries a name associated with large-scale musical performances and artistic complexity. It is designed for tasks that require sustained reasoning, nuance, and handling of lengthy, detailed content.

Within the Claude 3 generation, one model was built to prioritise speed and low cost without sacrificing too much quality. It is designed for high-volume tasks where quick responses matter more than deep reasoning. Its name reflects its lightweight nature, drawing from a small, delicate form in traditional craftsmanship.

One Claude release added a notable behaviour: before giving a final response, the model could work through a problem step by step internally. This extra reasoning stage improved performance on tasks involving logic, planning, and multi-part problems. It marked a shift towards more deliberate and structured problem-solving in the Claude family.