LLM-Powered CUDA Kernel Optimizer
Optimize CUDA kernels with an LLM-driven workflow.
Uses LM Studio by default for local execution
Provides live optimization feedback
Supports interactive code editing
Can be switched to OpenAI or Anthropic through
.env
Optimize
Configuration
❓ Help
Optimize
Configuration
❓ Help
Input
Load Example Kernel
Baseline CUDA Kernel
Model Provider
Max Iterations
↺
1
20
Start Optimization
Optimized Result
Best Optimized Code
Speedup Progress
Optimization Log