Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

FinanceAgent

A benchmark developed in collaboration with Stanford researchers, a Globally-Systemic Important Bank, and industry experts to evaluate AI agents' ability to perform tasks expected of an entry-level financial analyst. The dataset consists of 537 questions that assess various financial analysis skills.
Source:

Model Performance

#1
OpenAI-O3-medium
48.3%
#2
OpenAI-O3-high
48.3%
#3
Claude-4.1-Opus
46.1%
#4
Claude-4.0-Sonnet-Thinking
44.5%
#5
Claude-3.7-Sonnet-Thinking
44.1%
#6
Claude-4.0-Opus-Thinking
43.6%
#7
Claude-4.0-Sonnet
43.5%
#8
Claude-3.7-Sonnet
42.9%
#9
Claude-4.0-Opus
42.3%
#10
Grok-4
40.3%
#11
Grok-4-Thinking
40.3%
#12
OpenAI-O4-mini-medium
36.5%
#13
OpenAI-O4-mini-high
36.5%
#14
Grok-3-Mini-Beta
31.7%
#15
GPT-4.1
24.6%
#16
Grok-3-Beta
24.1%
#17
Grok-3-Think
24.1%
#18
OpenAI-O1-1217
20.8%
#19
Qwen-3-Max-Preview
17.7%
#20
Claude-3.5-Haiku
14.3%
#21
Gemini-2.0-Flash
13.2%
#22
OpenAI-O3-mini-high
12.7%
#23
OpenAI-O3-mini-medium
12.7%
#24
GPT-4-mini
10.8%
#25
Mistral-Medium-3.1
8.8%
#26
Magistral-Medium-3.1
7.5%
#27
Cohere-Command-A
4.3%
#28
Llama-4-Maverick-17B
3.6%
#29
Llama-3.3-70B
3.4%