$BTC
OpenAI Unveils GPT-5 — a Model That Thinks at a PhD Level
OpenAI has introduced GPT-5, a groundbreaking AI model that not only combines but surpasses the achievements of all previous versions. The model is more accurate, faster, and more logical, excelling in mathematics, writing, healthcare, and many other fields.
During the launch event, Sam Altman compared GPT-5 to “an expert with a PhD,” noting that GPT-3 was like a school student and GPT-4 like a university student. GPT-5 is designed for text creation, research, analysis, and programming — with coding receiving special attention. In the demo, GPT-5 built an educational website with an interactive game in just seconds. OpenAI calls it “the most powerful coding model to date.”
Benchmark scores for GPT-5:
⏺ 94.6% in AIME 2025 (Mathematics)
⏺ 74.9% in SWE-bench Verified (Programming)
⏺ 88% in Aider Polyglot (Programming)
⏺ 84.2% in MMMU (Multimodal Understanding)
⏺ 46.2% in HealthBench Hard (Healthcare)