Analysis
Invoice Payment-Fraud Detection
Given historical vendor records and a new invoice, flag suspicious bank-detail changes, duplicate invoice numbers, amount anomalies, and mismatched remittance instructions.
Finance & AccountingPer invoice
Not benchmarked
No result on this tier
Token profile pending
— /100K
Run this tier to add it
—
Granite 4.1 8B
8B · Off-the-shelf
832 in · 923 out · 12.53s p50
$13.39 /100K
Best current price · OpenRouter
Compare 2 router rates · observed 9/4/2026
OpenRouter · $0.0500 input · $0.1000 output /1M
TrustedRouter · $0.0527 input · $0.1055 output /1M
76%
Cheapest
Mistral Large 2407
123B · Off-the-shelf
1,106 in · 967 out · 24.14s p50
$211 /100K
Best current price · TrustedRouter
Compare 2 router rates · observed 9/11/2026
TrustedRouter · $0.5275 input · $1.58 output /1M
OpenRouter · $2.00 input · $6.00 output /1M
95%
GPT-5.4
Hosted · Off-the-shelf
828 in · 1,165 out · 11.07s p50
$1955 /100K
Best current price · OpenRouter
Compare 2 router rates · observed 9/11/2026
OpenRouter · $2.50 input · $15.00 output /1M
TrustedRouter · $2.64 input · $15.82 output /1M
100%
Best accuracy
All 12 model results and methodology
Generated 100/100 examples
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 72% | 100/100 | $42.21 |
| Qwen3 235B A22B | mid | 74% | 100/100 | $291 |
| DeepSeek V3 | mid | 81% | 100/100 | $107 |
| Mistral Large 2407 | mid | 95% | 100/100 | $211 |
| Arcee Trinity Large Thinking | mid | 55% | 100/100 | $140 |
| GPT-5.4 | frontier | 100% | 100/100 | $1955 |
| Claude Opus 4.7 | frontier | 73% | 100/100 | $5425 |
| Gemini 3.1 Pro Preview | frontier | 29% | 100/100 | $2530 |
| Gemma 4 E4B IT | small | 73% | 100/100 | $13.84 |
| Granite 4.1 8B | small | 76% | 100/100 | $13.39 |
| Ministral 8B Instruct 2410 | small | 68% | 100/100 | $36.24 |
| Qwen3 4B Instruct 2507 | small | 46% | 100/100 | $120 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-10T08:58:02.524Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock