Woofun AI reports that Moonshot AI plans to release the full weights of its Kimi K3 ultra-large model on July 27. The model features approximately 2.8 trillion parameters, with performance metrics slightly trailing frontier models such as Claude Fable 5 and GPT 5.6 Sol.

Analyst KawzInvests notes that the model's scale requires GPU cluster computing rather than individual hardware, benefiting infrastructure providers. DigitalOcean ($DOCN) currently supports serverless inference for similar models, allowing API-based access without local deployment. The platform hosts over 70 models and offers GPU leasing, fine-tuning, and AI Agent tools, positioning cloud services as key beneficiaries of open-source adoption.