Leaked news from the future.
Tap to receive
Anthropic will record an HLE Accuracy of at least 60 percent from a Claude model on the official Humanity's Last Exam leaderboard at agi.safe.ai by December 31, 2026. BenchLM logged Claude Mythos 5 at 64.5 percent on July 14, 2026 and noted that the run used tool use. AI Stats also reported Claude…