Create a Benchmark

Create a new benchmark run to measure model performance.

Step 1. Open New Benchmark Run

On the Evaluation → Benchmarks → Benchmark Run tab, click New Benchmark Run (top right) to open the run settings drawer.

---

Step 2. Fill in Settings

Field	Required	Description
Serverless Endpoint	✓	A running Serverless endpoint to evaluate. If none exists, create and start one in AI Inference first.
Local Queue Selection	✓	Local queue to run on (default).

---

Step 3. Run

Click Run to start the benchmark run. Cancel aborts.

Run multiple benchmarks against the same endpoint to track performance changes.