FeatureFeb 13, 2026
Hugging FaceCustom Kernels for All from Codex and Claude
Why It Matters
This release simplifies the complex process of writing and integrating CUDA kernels, making it accessible for developers to optimize AI models efficiently.
Release Summary
Hugging Face introduces an agent skill for coding agents to write CUDA kernels.
The skill integrates with Codex and Claude to produce working kernels for diffusers and transformers.
It includes architecture-specific optimization guides and PyTorch binding instructions.
The skill is available through the kernels library and can be installed with a single command.
This entry is based on publicly available announcements. AI Product Release Radar is not affiliated with Hugging Face. No guarantee of accuracy. Not financial advice.
AD_SLOT