Skip to content

Benchmark

FinBen MultiFin-ES ACC

Scores are percentages. Higher is better.

Models measured
13
Scores recorded
13
Sources
1
IQ weight
1.20

Score distribution

36.5%
Mean
37.8%
Median
16%
Std dev
13%
10th percentile
60%
90th percentile
6.1%
Min
6.1%61.3%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.20 for chat models, measured over 13 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1DeepSeek V3published-demo-leaderboard, MultiFin-ES, ACC· TheFinAI public numeric results61.3%
2GPT-4 Turbopublished-demo-leaderboard, MultiFin-ES, ACC· TheFinAI public numeric results60%
3GPT-3.5 Turbopublished-demo-leaderboard, MultiFin-ES, ACC· TheFinAI public numeric results48%
4Qwen2.5 7B Instructpublished-demo-leaderboard, MultiFin-ES, Acc· TheFinAI public numeric results46.5%
5Qwen2 7B Instructpublished-demo-leaderboard, MultiFin-ES, ACC· TheFinAI public numeric results43%
6Llama 3.2 3Bpublished-demo-leaderboard, MultiFin-ES, Acc· TheFinAI public numeric results40.4%
7codellama-7bpublished-demo-leaderboard, MultiFin-ES, Acc· TheFinAI public numeric results37.8%
8Mistral (7B) Instruct v0.3published-demo-leaderboard, MultiFin-ES, Acc· TheFinAI public numeric results36.5%
9Qwen2.5 7Bpublished-demo-leaderboard, MultiFin-ES, Acc· TheFinAI public numeric results36.1%
10Llama-2-7b-chat-hfpublished-demo-leaderboard, MultiFin-ES, ACC· TheFinAI public numeric results23%
11DeepSeek R1 Distill Qwen 7Bpublished-demo-leaderboard, MultiFin-ES, ACC· TheFinAI public numeric results22.2%
12Mixtral 8X7b V0.1published-demo-leaderboard, MultiFin-ES, ACC· TheFinAI public numeric results13%
13code-llama-7b-instructpublished-demo-leaderboard, MultiFin-ES, Acc· TheFinAI public numeric results6.1%

Data credits: The Fin AI. Published FinBen result summaries and Open Greek Financial LLM Leaderboard aggregates. Results licence not separately stated; fractional FinBen success metrics converted to percentages.