Back to Blog
Tag: Moonshot
3 posts found
Complete guide to Kimi K2.5: 1T parameter MoE model with 32B active, MIT license, native multimodal architecture. Covers Agent Swarm with 100 parallel agents, PARL training, vision coding capabilities, benchmarks vs GPT-5.2/Claude 4.5/Gemini 3, hardware requirements, API pricing, and deployment.
Comprehensive guide to Kimi K2 Thinking: Moonshot AI's deep reasoning model with chain-of-thought visualization. Learn API integration, best practices, and optimization strategies.
Kimi K2-0905 is Moonshot AI's latest open-source large language model featuring 1 trillion parameters with 32B active, 256K context length, and exceptional coding capabilities approaching Claude Sonnet 4 level.