Woofun AI reports that The Dark Side of the Moon has officially launched the Kimi K3 flagship model, featuring 2.8 trillion parameters and a hybrid expert architecture. The system supports a 1 million token context window and processes text, images, and videos simultaneously, targeting programming, research, and knowledge work. Each inference activates only 16 of 896 experts. Official benchmarks indicate K3 surpasses Claude Opus 4.8 and GPT-5.5 in specific programming and Agent tasks, though it trails behind Claude Fable 5 and GPT-5.6 Sol overall. The model is available via Kimi, Kimi Work, Kimi Code, and API, with input priced at $3 and output at $15 per million tokens. Full model weights are scheduled for open-source release no later than July 27.