The article summarizes key developments in large language models (LLMs) over the past six months, highlighting the intense competition between major AI providers including OpenAI, Anthropic, and Google. It focuses on the 'November 2025 inflection point' when coding capabilities saw significant improvements, with models like GPT-5.1, Gemini 3, and Claude Opus 4.5 rapidly surpassing each other. The author uses a unique 'pelican riding a bicycle' test to illustrate model differences and notes the emergence of more capable coding agents through Reinforcement Learning from Verifiable Rewards.
Background
Large language models have been rapidly evolving, with major tech companies competing to develop the most capable models, particularly in specialized areas like coding. The field has seen frequent model updates and improvements in capabilities.
- Source
- Simon Willison
- Published
- May 19, 2026 at 09:09 AM
- Score
- 6.0 / 10