Domestic Open-Source World-Class Model: What Is the Strength of Kimi K3?
Moonshot AI
Moonshot AI (月之暗面) has released its new open-source large language model, Kimi K3. The model achieves impressive results on international benchmarks: it ranks first on LiveCodeBench (programming tasks), surpassing GPT-4o and Claude 3.5 Sonnet, and also demonstrates high scores on AIME 2025 (mathematics) and MMLU-Pro (general knowledge).
Chinese startup Moonshot AI (月之暗面) has released a new open-source large language model called Kimi K3. The model achieved outstanding results on several international benchmarks. In the LiveCodeBench test, which evaluates programming skills, Kimi K3 ranked first, surpassing well-known models such as OpenAI's GPT-4o and Anthropic's Claude 3.5 Sonnet. Additionally, the model demonstrated high performance on the AIME 2025 exam (mathematics) and the MMLU-Pro test (general knowledge). Developers note that Kimi K3 combines cutting-edge technology with accessibility through its open-source nature, which could facilitate its widespread adoption both in China and abroad.
Source: Moonshot Kimi (GNews) —
original