[ignoramous]

> continuing to refuse to use chinese models due to security and IP concerns

One can run open weights in an exclusive TEE'd GPU too, which still comes out cheaper than closed weight LLMs. Ex: https://chutes.ai/pricing