Qwen 3.5 4B AI Pod runtimes
Docker save.gz archives for the AI Pod Qwen 3.5 4B application. These contain the inference runtime and dependencies. Download the GGUF model and vision projector separately from unsloth/Qwen3.5-4B-MTP-GGUF.
| Device | Archive | Bytes | SHA256 |
|---|---|---|---|
| orin | runtime-orin-0.1.2.tar.gz |
319156055 | b5305b5a6e4988ab13acae888f515602daeab4ae0e820a7d4fb264115b17eaa3 |
| thor | runtime-thor-0.1.1.tar.gz |
777448570 | a2eb912cf6096d0d7a660df8feada8c1ea680817506a68f9228a8a6b70b143c5 |
AGX Orin uses runtime 0.1.2 with CUDA VMM disabled; AGX Thor uses runtime 0.1.1. Extract with docker load -i <archive>. The application checks archive size, SHA256 and loaded image identity.
Licenses for llama.cpp and NVIDIA CUDA/NCCL are included in licenses/. Each component remains subject to its own license.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support