# ScreenSpot Pro with Python (ScreenSpot Pro w/ Python)

> ScreenSpot Pro GUI grounding on high-resolution professional screenshots, with the model allowed to run Python (for example to crop or zoom) before answering.

Canonical page: https://benchlm.ai/benchmarks/screenspotpropython

- Category: [Multimodal & Grounded](/multimodal-grounded)
- Last updated: November 13, 2026

## About ScreenSpot Pro w/ Python

- Year: 2026
- Tasks: 1,581 GUI-grounding instructions on professional screenshots
- Format: Accuracy (%)
- Difficulty: High-resolution GUI grounding with tools
- Paper: [Introducing Inkling 2 Small](https://thinkingmachines.ai/news/inkling-2-small/)

Thinking Machines reports the with-Python setting next to the standard ScreenSpot Pro score in the Inkling 2 Small launch (November 5, 2026): Inkling 2 Small 91.5% (83.8% without tools), GPT-6 Luna 87.6% and Kimi K3 83.2%. The no-tools score stays on the ScreenSpot Pro page.

ScreenSpot Pro w/ Python is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (1 models)

The one result on this page was reported by the lab that built the model.

| Rank | Model | Creator | Score | Who ran it |
|------|-------|---------|-------|-----------|
| 1 | [Inkling 2 Small](/models/inkling-2-small) | Thinking Machines Lab | 91.5% | Reported by Thinking Machines Lab |

## FAQ

### What does ScreenSpot Pro w/ Python measure?

ScreenSpot Pro GUI grounding on high-resolution professional screenshots, with the model allowed to run Python (for example to crop or zoom) before answering.

### Which model scores highest on ScreenSpot Pro w/ Python?

Inkling 2 Small by Thinking Machines Lab currently leads with a score of 91.5% on ScreenSpot Pro w/ Python.

### How many models are evaluated on ScreenSpot Pro w/ Python?

1 AI models have published results on ScreenSpot Pro w/ Python in the BenchLM catalog.

### Does ScreenSpot Pro w/ Python affect BenchLM's overall score?

Not directly. ScreenSpot Pro w/ Python is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.