According to @DeepLearningAI, Moonshot AI launched Kimi K2 Thinking and Kimi K2 Thinking Turbo that alternate between cycles of reasoning and tool use, often making hundreds of calls, and they outperform other open-weights LLMs on complex, multi-step tasks. Source: DeepLearning.AI (Nov 22, 2025). According to @DeepLearningAI, both models are trillion-parameter mixture-of-experts systems fine-tuned at INT4 precision, delivering strong agentic performance while running on lower-cost hardware. Source: DeepLearning.AI (Nov 22, 2025). According to @DeepLearningAI, key trading-relevant datapoints are the multi-call agent workflow and INT4 efficiency on cheaper hardware, which directly inform cost and throughput assumptions for AI infrastructure exposure across traditional and crypto markets. Source: DeepLearning.AI (Nov 22, 2025).

According to DeepLearningAI, Beijing-based Moonshot AI has released the Kimi K2 large language model (LLM) family, offering open-weights access under a modified MIT license to a one trillion-parameter model. The fine-tuned Kimi-K2-Instruct version achieved 53 percent on LiveCodeBench and 76.5 percent on AceBench, outperforming other models in these benchmarks. This open release is expected to accelerate AI-driven innovation and could significantly impact crypto markets as more projects leverage powerful, accessible AI for DeFi, trading bots, and blockchain analytics (source: DeepLearningAI).