Tencent has released the weights of its Hy4 model while imposing restrictions on large hosting providers, allowing commercial use and modification, though qualifying model service operators must obtain a separate license. The default deployment for vLLM utilizes eight B300 GPUs.