China’s Groundbreaking Language Model ‘KIMI K2’ Unveiled on the 11th, Captivating Global Interest

Chinese AI start-up Moonshot AI has launched its innovative model ‘Kimi K2,’ a large language model (LLM) featuring approximately 1 trillion parameters. Launched in its third year, Kimi K2 has already eclipsed benchmarks set by competitors like DeepSeek, showcasing exceptional coding capabilities. Utilizing a “Mixture of Experts” (MoE) method, Kimi K2 activates only 32 billion parameters for inference tasks, optimizing performance and cost-effectiveness. It achieved notable results on coding benchmarks, ranking second in SWE-Bench Verified and leading in LiveCodeBench against major players like OpenAI. Recognized for its potential, Moonshot AI has become a unicorn, valued at $2.5 billion shortly after its founding in 2023. The company has since attracted significant investments from Alibaba and Tencent, raising its valuation to $3.3 billion. Led by Dr. Yang Jilin, a Tsinghua University alumnus, Moonshot AI aims to advance towards artificial general intelligence (AGI).

Source link