OpenAI's latest GPT-5.4 model, released March 5, 2026, has propelled its Humanity's Last Exam score to 41.6% without tools—up sharply from GPT-5.2's 35% in January—reflecting rapid scaling in reasoning across 2,500 expert-level questions designed to resist benchmark saturation. This positions OpenAI near the top of leaderboards, trailing Google's Gemini 3 Deep Think at 48.4%, amid intensifying competition from Anthropic's Claude Opus 4.6 and Chinese labs like Moonshot's Kimi-K2. With tools, GPT-5.4 exceeds 50%, but the no-tools metric defines frontier capabilities. Traders eye potential GPT-5.5 or o5 releases by June 30, alongside developer conferences and funding rounds that could accelerate progress or reveal delays in AI safety evaluations.
Экспериментальная сводка, созданная ИИ на основе данных Polymarket · ОбновленоОценка OpenAI GPT на последнем экзамене человечества к 30 июня?
Оценка OpenAI GPT на последнем экзамене человечества к 30 июня?
50%+
31%
$0.00 Объем
50%+
31%
The resolution source will be the official Humanity’s Last Exam leaderboard https://scale.com/leaderboard/humanitys_last_exam.
Открытие рынка: Jan 30, 2026, 12:00 AM ET
Resolver
0x65070BE91...The resolution source will be the official Humanity’s Last Exam leaderboard https://scale.com/leaderboard/humanitys_last_exam.
Resolver
0x65070BE91...OpenAI's latest GPT-5.4 model, released March 5, 2026, has propelled its Humanity's Last Exam score to 41.6% without tools—up sharply from GPT-5.2's 35% in January—reflecting rapid scaling in reasoning across 2,500 expert-level questions designed to resist benchmark saturation. This positions OpenAI near the top of leaderboards, trailing Google's Gemini 3 Deep Think at 48.4%, amid intensifying competition from Anthropic's Claude Opus 4.6 and Chinese labs like Moonshot's Kimi-K2. With tools, GPT-5.4 exceeds 50%, but the no-tools metric defines frontier capabilities. Traders eye potential GPT-5.5 or o5 releases by June 30, alongside developer conferences and funding rounds that could accelerate progress or reveal delays in AI safety evaluations.
Экспериментальная сводка, созданная ИИ на основе данных Polymarket · Обновлено
Не доверяй внешним ссылкам.
Не доверяй внешним ссылкам.
Часто задаваемые вопросы