Skip to content

Benchmark

FinBen ccfraud F1

Scores are percentages. Higher is better.

Models measured
15
Scores recorded
15
Sources
1
IQ weight
0.21

Score distribution

51.8%
Mean
48%
Median
29.5%
Std dev
17%
10th percentile
91.3%
90th percentile
0.65%
Min
0.65%92.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.21 for chat models, measured over 15 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Qwen2.5 7Bpublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results92.2%
2DeepSeek R1 Distill Qwen 7Bpublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results91.3%
3gemini-1.0-propublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results90%
4Qwen2.5 7B Instructpublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results85.9%
5codellama-7bpublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results79%
6DeepSeek V3published-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results61.2%
7GPT-4 Turbopublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results55%
8Mixtral 8X7b V0.1published-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results48%
9code-llama-7b-instructpublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results43.8%
10Qwen2 7B Instructpublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results42%
11Mistral (7B) Instruct v0.3published-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results25.9%
12Llama-2-7b-chat-hfpublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results25%
13GPT-3.5 Turbopublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results20%
14Llama 2 70bpublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results17%
15Llama 3.2 3Bpublished-demo-leaderboard, ccfraud, F1· TheFinAI public numeric results0.65%

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.