Submit
Your coding agent can measure how fast a model runs on your machine and send the result straight to llamaperf. It works with Claude Code, Cursor, Claude Desktop and any other client that speaks MCP.
Let your agent run the benchmark
Ask your agent to benchmark the model you use. It runs the test on your machine, reads your GPU or Mac and its memory, shows you the numbers and sends them as a report once you agree.
The agent connects with a key tied to your account, so you need one first. It's free.