Grounding scenario · 2026-08-13

Locate the Add device button in Smart home

Every model received the same screenshot, prompt, and structured response contract.

Case IDadd device

390 × 844 · normalized 0–1000 coordinates

Exact input

Prompt

You are evaluating a user interface screenshot for test automation.

Locate the single UI element described below. Return the best click point and the tight visible bounding box of the complete interactive target. Coordinates use a normalized 0 to 1000 grid, where (0, 0) is the top-left image corner and (1000, 1000) is the bottom-right image corner.

Target: the Add device button

Return only data matching the supplied JSON schema. Do not describe your reasoning.
SHA-256 · 04600212f2b6f7aab1cd0c261d803087507d642468b07b6775f585c26e721c7e
Smart home UI with the the Add device button target highlightedExpected
Ground truth: the Add device button

Model evidence

9 responses

Blue marks the browser-derived target. Orange marks the model’s returned box and point. Format validity checks only the response contract; it does not mean the requested element was found.

Holo3 122B A10B

H Company · holo3-122b-a10b
Point hit
Format
Valid
Box IoU
0.881
Latency
1.47 s
Cost
$0.00034
Raw response
{"point": {"x": 500,  	"y": 877}, "box": {"x": 43, "y": 850, "width": 957, "height": 54}}

Grok 4.5

xAI · x-ai/grok-4.5
Point hit
Format
Valid
Box IoU
0.654
Latency
1.53 s
Cost
$0.0018
Raw response
{"point":{"x":500,"y":865},"box":{"x":50,"y":825,"width":900,"height":80}}

GPT-5.4

OpenAI · openai/gpt-5.4
Point hit
Format
Valid
Box IoU
0.919
Latency
1.50 s
Cost
$0.0025
Raw response
{"point":{"x":500,"y":877},"box":{"x":18,"y":851,"width":939,"height":54}}

Gemini 3.6 Flash

Google AI Studio · google/gemini-3.6-flash
Point hit
Format
Valid
Box IoU
0.977
Latency
1.77 s
Cost
$0.0022
Raw response
{"point": {"x": 500, "y": 878}, "box": {"x": 46, "y": 850, "width": 908, "height": 57}}

GPT-5.6 Sol

OpenAI · openai/gpt-5.6-sol
Point hit
Format
Valid
Box IoU
0.589
Latency
2.30 s
Cost
$0.0046
Raw response
{"point":{"x":500,"y":894},"box":{"x":46,"y":865,"width":908,"height":58}}

GPT-5.6 Luna

OpenAI · openai/gpt-5.6-luna
Point hit
Format
Valid
Box IoU
0.960
Latency
1.06 s
Cost
$0.00009
Raw response
{"point":{"x":500,"y":879},"box":{"x":46,"y":850,"width":908,"height":56}}

Claude Sonnet 5

Anthropic · anthropic/claude-sonnet-5
Point miss
Format
Valid
Box IoU
0.000
Latency
4.27 s
Cost
$0.0028
Raw response
{"point":{"x":195,"y":742},"box":{"x":18,"y":719,"width":354,"height":46}}

Claude Opus 5

Anthropic · anthropic/claude-opus-5
Point miss
Format
Invalid
Box IoU
0.000
Latency
3.89 s
Cost
$0.0071
Raw response
{"point":{"x":500,"y":1904},"box":{"x":43,"y":1848,"width":915,"height":112}}

point must use normalized 0..1000 coordinates. box must be positive and contained in normalized 0..1000 coordinates.

Gemini 3.1 Pro Preview

Google AI Studio · google/gemini-3.1-pro-preview
Point hit
Format
Valid
Box IoU
0.977
Latency
3.35 s
Cost
$0.0030
Raw response
{"point": {"x": 500, "y": 878}, "box": {"x": 46, "y": 850, "width": 908, "height": 57}}
EN