TensorIR

license

Apache-2.0(OSI)

source

public

core-gated

ungated

Apache-2.0 license body confirmed. The repository is public and unarchived and builds the whole product, and the README describes no paid tier, enterprise edition or license-gated build beside it, so source is public and the core ungated.

• https://github.com/NVIDIA/tensor-ir/blob/main/LICENSE recorded 2026-09-02

  LICENSE body carries the verbatim Apache License 2.0 text under an NVIDIA copyright header.

• https://api.github.com/repos/NVIDIA/tensor-ir recorded 2026-09-02

  Repo metadata for nvidia-tensor-ir's canonical repository - private false, archived false, fork false.

• https://github.com/NVIDIA/tensor-ir/blob/main/README.md recorded 2026-09-02

  README describes TensorIR Compiler as an early release for community evaluation that depends on CUDA Tile IR for GPU code generation, and names no paid tier, enterprise edition or license-gated build beside the published source.

111 GitHub stars, which lands in the <1K stars band of the stars scale, level 1. No package registry publishes Tensor IR and the repository is built from source, so no download count exists and the stars scale applies.

• https://api.github.com/repos/NVIDIA/tensor-ir recorded 2026-09-02

  Repo metadata - stargazers_count = 111 - for NVIDIA/tensor-ir.

One band below the apache-tvm anchor, on the same reasoning as cuda-tile, the backend it depends on: a full compilation toolchain narrowed to one vendor's targets (NVIDIA GPUs) rather than apache-tvm's multi-vendor retargetability.

• https://github.com/NVIDIA/tensor-ir/blob/main/README.md recorded 2026-09-02

  README describes TensorIR Compiler as an early release for community evaluation that depends on CUDA Tile IR for GPU code generation, and names no paid tier, enterprise edition or license-gated build beside the published source.