Published on February 22, 2026, the entry describes a C++/CUDA inference engine that claims to run Llama 70B on an RTX 3090, transferring data from NVMe to the GPU without passing through the CPU.
The February 22, 2026 entry presents ntransformer as a C++/CUDA inference engine for language models. According to the project page description, it runs Llama 70B on an RTX 3090 and transfers data from NVMe to the GPU without passing through the CPU.
The claim may interest engineers evaluating large-model execution on a single consumer GPU. To check the result, consult the original project page and review its instructions, hardware requirements, and available test methods; this entry provides no performance figures or further configuration details.