Classification
FCPA Expense Interaction Risk Classification
Given synthetic expense records and interaction notes, classify potential foreign-official risk using a supplied taxonomy and cite the triggering fields.
Legal & CompliancePer expense record
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
501 in · 60 out · 0.93s p50
$3.10 /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
60%
Qwen3 235B A22B
235B (22B active) · Off-the-shelf
515 in · 61 out · 1.14s p50
$34.53 /100K
Best current price · OpenRouter
Compare 2 router rates · observed 9/11/2026
OpenRouter · $0.4550 input · $1.82 output /1M
TrustedRouter · $0.6330 input · $2.11 output /1M
90%
GPT-5.4
Hosted · Off-the-shelf
494 in · 97 out · 1.80s p50
$269 /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
97%
Best accuracy
All 12 model results and methodology
Generated 100/100 examples
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 58% | 100/100 | $13.68 |
| Qwen3 235B A22B | mid | 90% | 100/100 | $34.53 |
| DeepSeek V3 | mid | 74% | 100/100 | $18.75 |
| Mistral Large 2407 | mid | 85% | 100/100 | $41.62 |
| Arcee Trinity Large Thinking | mid | 73% | 100/100 | $108 |
| GPT-5.4 | frontier | 97% | 100/100 | $269 |
| Claude Opus 4.7 | frontier | 96% | 100/100 | $722 |
| Gemini 3.1 Pro Preview | frontier | 92% | 100/100 | $1003 |
| Gemma 4 E4B IT | small | 58% | 100/100 | $1.43 |
| Granite 4.1 8B | small | 60% | 100/100 | $3.10 |
| Ministral 8B Instruct 2410 | small | 48% | 100/100 | $9.02 |
| Qwen3 4B Instruct 2507 | small | 58% | 100/100 | $15.87 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-08T19:27:03.014Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock