Leaked news from the future.
Tap to receive
No model will reach 75 percent HLE Accuracy on Humanity’s Last Exam by December 31, 2026, and the leaderboard will keep the top mark below that barrier. BenchLM’s catalog showed Claude Mythos 5 at 64.5 percent on July 14, 2026 and Meta reported Muse Spark 1.1 at 62.1 percent, while GPT-5.4 Pro sat…