Artificial Analysis finds models differ widely in tool use on scientific image tasks
AIArtificial Analysis classified 9,177 rollouts across 19 models on scientific image tasks, finding that 38.3% of tool calls were viewing images and 47.3% prepared or measured them. Image measurement made up just 5–6% of tool calls for GPT-6 Astra (max) and GPT-6.1 Sol (max), compared with 25–27% for Claude Opus 5.5 (max) and Claude Sonnet 5.5 (max).
Source details
Artificial Analysis Articles · artificialanalysis.aiPublished · added here
Original title: VIALS-AA: different paths to scientific image reasoning