TMTPOST – Kimi has launched its visual thinking model, K1, which is built on reinforcement learning technology. The K1 model natively supports end-to-end image understanding and chain-of-thought techniques, extending its capabilities beyond mathematics to a broader range of foundational science fields. In benchmark tests for basic science disciplines such as mathematics, physics, and chemistry, the initial K1 model outperformed global benchmark models, including OpenAI’s O1, GPT-4, and Claude 3.5 Sonnet.