Last year we built megakernels. This year we deleted them: The road to Together Compiler
Imported from official source
Together Compiler uses AI agents to rewrite and fuse the GPU kernels in a model’s forward pass. It beats hand-written megakernels by 28% on B200 (2,118 vs 1,649 tokens/s on Llama 3.2 1B). This is how we got from one to the other.
This version
- Version
- 1 of 1
- Recorded
- November 16, 2026 23:30
- Change
- Initial
- Content hash
ac83a1e0facddff804f776527bbd1707- All versions
- Revision history