What's new on Artificial Analysis
artificialanalysis.ai ↗Table view
Coding agents — harness + model (Artificial Analysis Coding Agent Index)
The same model scores differently depending on the agent driving it. Each row is one agent running one model on DeepSWE, SWE-Atlas-QnA and Terminal-Bench v4; cost, time and tokens are per task. Click a header to sort. Source ↗