Frontierbench.ai describes the project as "Terminal-Bench 3.0 (formerly Frontier-Bench)," meaning it supersedes rather than competes with Terminal-Bench 2.x — model cards that mention Frontier-Bench have so far also reported Terminal-Bench in the same document. Its task set is not comparable to earlier Terminal-Bench versions, and it should not be confused with Cognition's separately maintained FrontierCode benchmark despite the similar name.