benchmark evidence
ScreenSpot-Pro
ScreenSpot-Pro GUI grounding micro-average accuracy.
activeupstream source ->
winner on ScreenSpot-Pro
direct benchmark result, not a broad vertical composite | source row dated 2026-04-14
scored on 2026-04-14 · stale source data (46d)
latest mapped results | top 20
| # | Model | Score | Evidence | Tested |
|---|---|---|---|---|
| 1 | Anthropic: Claude Haiku 4.5 | 17.0 | model-only independent_benchmark | 2026-04-14 |
| 2 | OpenAI: GPT-5 | 0.8 | model-only independent_benchmark | 2026-04-14 |
what this result means
ScreenSpot-Pro GUI grounding micro-average accuracy.
This benchmark contributes direct public evidence. Read its scope before generalizing the result.
A win here is a win on ScreenSpot-Pro. Broad task pages require independent corroboration before naming a general winner.
source record
category: computer_use
metric: accuracy
matched models: 2
latest source date: 2026-04-14
direction: higher is better