Backend Bench
Environment to evaluate LLMs on the ability to generate correct and fast GPU kernels, passing tests provided by `Torch`
- Domain
- rl-env
- License
- unknown
- Published
- Sep 2025
Cite
Notes
Only stored in your browser.
Top score 5.0% by Claude Sonnet 4.5 - 1 model reporting (1 frontier)
Top models
1
Related tools
1Implementations, trainers, datasets and scaffolds linked to this eval.
FAQ
- What is Backend Bench?
- Environment to evaluate LLMs on the ability to generate correct and fast GPU kernels, passing tests provided by `Torch`
- What is the current top score on Backend Bench?
- The top reported score is 5.0% by Claude Sonnet 4.5, across 1 model reporting (1 from frontier labs).
- How can a model improve its Backend Bench score?
- Tools linked to Backend Bench on Sophon include Backend Bench RL Env (Community) - RL environments, datasets, and scaffolds that target this eval.
- What license is Backend Bench under?
- Backend Bench is available under unknown.