Will an open-weight model match or be within 2 points of the top closed model on a major coding benchmark before 2027-12-31?

Open-weight models close the gap

Fast open-source iteration, cheap training-method innovation, and hosted inference removing the local-hardware wall let open-weight models track closed frontier releases within months. Developers and researchers describe switching workflows to hosted open-weight endpoints without losing capability, and note that competitive open-weight releases intensify pressure on closed labs. These dynamics point toward an open-weight model reaching within 2 points of the top closed model on a headline public coding benchmark before 2027, undermining the claim that frontier capability is a defensible closed-lab moat.

6 voices