Stagehand Agent Evals

See how your favorite models perform on various computer use benchmarks. Compare across accuracy, cost, and speed.

Accuracy

Rank	Model	Provider	Accuracy (%)	Cost/Task ($)	Speed (S)
1	claude-fable-5	Anthropic	90.62	0.522	530.00
2	claude-opus-4-8	Anthropic	83.33	0.448	184.68
3	gpt-5.5	OpenAI	76.19	1.275	149.18
4	gemini-3-flash-preview	Google	73.81	0.029	112.57
5	claude-opus-4-7	Anthropic	69.05	0.768	116.45
6	claude-sonnet-4-6	Anthropic	66.67	0.483	161.11
7	claude-haiku-4-5	Anthropic	61.90	0.172	182.45
8	claude-sonnet-4-5	Anthropic	59.52	0.573	215.41
9	gemini-2.5-computer-use-preview-10-2025	Google	42.86	0.084	161.20
10	gpt-5.4	OpenAI	42.86	0.397	229.29
11	gpt-5.4-mini	OpenAI	38.10	0.058	67.05

Top performers

Top accuracy

90.6%

claude-fable-5

Lowest cost

$0.0289

gemini-3-flash-preview

Fastest speed

67.05s

gpt-5.4-mini

We can help you train and evaluate your models.

Get a custom evaluation of your models on Stagehand and Browserbase.

Get in touch Read the docs

Rank

Model

Provider

Accuracy (%)

Cost/Task ($)

Speed (S)

claude-fable-5

Anthropic

90.62

0.522

530.00

claude-opus-4-8

Anthropic

83.33

0.448

184.68

gpt-5.5

OpenAI

76.19

1.275

149.18

gemini-3-flash-preview

Google

73.81

0.029

112.57

claude-opus-4-7

Anthropic

69.05

0.768

116.45

claude-sonnet-4-6

Anthropic

66.67

0.483

161.11

claude-haiku-4-5

Anthropic

61.90

0.172

182.45

claude-sonnet-4-5

Anthropic

59.52

0.573

215.41

gemini-2.5-computer-use-preview-10-2025

Google

42.86

0.084

161.20

gpt-5.4

OpenAI

42.86

0.397

229.29

gpt-5.4-mini

OpenAI

38.10

0.058

67.05