Artificial Analysis releases VIALS-AA scientific image reasoning benchmark
Understand this story
Artificial Analysis has released the VIALS-AA scientific image reasoning benchmark, analyzing 9,177 rollouts of 19 models on scientific image reasoning tasks and categorizing the tool calls used by these models. The statistics show that 38.3% of tool calls were used to view images, and 47.3% were used for image preparation and measurement operations such as cropping, enhancement, and pixel measurement.
AI overview based on source reports · Updated 1 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Artificial Analysis ArticlesVIALS-AA: different paths to scientific image reasoning
VIALS-AA analyzes how 19 models reason over scientific images, classifying 9,177 tool calls from its original test set: 38.3% were image viewing and 47.3% prepared or measured images. Image measurement made up just 5–6% of tool calls for GPT-6 Astra (max) and GPT-6.1 Sol (max), versus 25–27% for Claude Opus 5.5 (max) and Claude Sonnet 5.5 (max). The report also notes that Gemini 3.8 Flash (high) accounts for 86.7% of commentary/no-op calls, that more intelligent models tend to call fewer tools, and that cost per task varies by nearly 100× across models.
Activity for this story
Not enough continuous observations to show a trend yet.