A pull request adds support for the Eagle3 model in Qwen3.5 and Qwen3.6 within llama.cpp. The change is proposed by ruixiang63 and submitted to the ggml-org/llama.cpp repository.