Five models that mattered, ranked by influence rather than raw benchmark score. The ordering is opinionated; the criteria are not.
- The one that proved scaling laws work.
- The one that made instruction-tuning the default.
- The one that opened the door to multimodal interaction.
- The one that finally made agentic workflows reliable.
- The one quietly running in production while everyone argued about the rest.
We won’t name them — half the fun is the argument.