New
hardenedlinux/sonicboom v0.0.2 — a CUDA tensor runtime without Python
hardenedlinux/sonicboom is a native tensor runtime built in C++ and CUDA; in the README's limited comparison, decoding was faster than llama.cpp while prefill was slower.
SonicBoom is an AI framework written from the ground up in C++ and CUDA, for workloads where performance matters. Everything is exposed through a stable C ABI, and GNU Guile is its first-class language.
View on GitHub ↗hardenedlinux/sonicboom is a native tensor runtime built in C++ and CUDA; in the README's limited comparison, decoding was faster than llama.cpp while prefill was slower.
No matching stories.