Opus 5 vs GPT-5.6 Sol vs Kimi K3: Who Leads Now?

The article discusses the latest flagship models from OpenAI, Moonshot, and Anthropic, specifically GPT-5.6 Sol, Kimi K3, and Claude Opus 5. Opus 5 leads in SWE-bench Pro and ARC-AGI-3, while Sol holds Terminal-Bench 2.1. Kimi K3 is a 2.8 trillion parameter open-weight model with a lower input price. The models have converged in specs, and the differences lie in behavior under load. Engineers should note the differences in behavior and consider the specific use case when choosing a model.