KingBench3
A proprietary programming evaluation benchmark that judges model outputs by whether a user can practically use the produced artifact in a real browser or runtime environment, distinguishing it from conventional code generation benchmarks based on unit test pass rates.
Timeline (last 90 days)
KingBench3 is a private benchmark designed by the tester, with the sole evaluation criterion being whether the output can actually be used by a user in a real browser/runtime environment, rather than whether the code passes unit tests.
Le Chonk scored a total of 42/80, or 52.5%, equivalent to 5.25/10, across the eight questions in KingBench3.
All Facts (2)
KingBench3 is a private benchmark designed by the tester, with the sole evaluation criterion being whether the output can actually be used by a user in a real browser/runtime environment, rather than whether the code passes unit tests.
50%UnverifiedLe Chonk scored a total of 42/80, or 52.5%, equivalent to 5.25/10, across the eight questions in KingBench3.
50%