GPT-5

GPT-5, OpenAI's multimodal flagship launched August 7, 2025, deploys Mixture of Experts and Chain-of-Thought Prompting, but its edge is eroding fast. Anthropic's Claude Opus 4.7 just posted 80.1% on SWE-Bench with 1M context, while AllenAI's MolmoAct2 beat GPT-5 on robotics with a 720-hour bimanual dataset. Both models explicitly list GPT-5 as a competitor. Meanwhile, OpenAI shipped GPT-5.5 in April 2026 targeting smarter agents and deeper tool use, suggesting an accelerated cadence driven by competitive pressure. A recent study also flagged GPT-5's central tendency bias in clinical scoring, a reliability concern for high-stakes deployments. The model now powers ChatGPT and Microsoft's HORIZON product, widening its distribution surface but also its liability.

• ·GPT-5 uses Sparse MoE and Chain-of-Thought, launched August 2025
• ·Competes directly with Claude Opus 4.7 (80.1% SWE-Bench) and MolmoAct2 (robotics leader)
• ·GPT-5.5 shipped April 2026, indicating accelerated iteration cycle
• ·Central tendency bias in clinical scoring flagged in recent study
• ·Distributed via ChatGPT and Microsoft HORIZON product