Framework & Tool
JAX
Google's high-performance numerical-computing library with NumPy-like API and XLA backend.
Definition
JAX combines a NumPy-compatible API with function transforms (jit, vmap, grad, pmap) and XLA compilation. Google uses it to train Gemini and most internal models. It also underpins Flax, Haiku, and the Equinox ecosystem of research frameworks.
Common use cases
- Research
- Large-scale training
- TPU workloads