Classification
Expense Policy Decision Explanation
Given an expense claim and the applicable policy, classify it as approve, reject, or review and return the exact policy rule supporting the decision.
Finance & AccountingPer expense claim
Not benchmarked
No result on this tier
Token profile pending
— /100K
Run this tier to add it
—
Qwen3 4B Instruct 2507
4B · Off-the-shelf
322 in · 81 out · 2.96s p50
$9.98 /100K
AWS Bedrock temporary proxy price
View token rates
AWS Bedrock · $0.1545 input · $0.6180 output /1M
31%
DeepSeek V3
Hosted · Off-the-shelf
304 in · 32 out · 0.77s p50
$12.46 /100K
Best current price · OpenRouter
Compare 2 router rates · observed 9/11/2026
OpenRouter · $0.2900 input · $1.14 output /1M
TrustedRouter · $0.3376 input · $0.9390 output /1M
40%
GPT-5.4
Hosted · Off-the-shelf
299 in · 76 out · 1.74s p50
$189 /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
83%
Best accuracy
All 12 model results and methodology
Generated 100/100 examples
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 18% | 100/100 | $13.44 |
| Qwen3 235B A22B | mid | 24% | 100/100 | $25.21 |
| DeepSeek V3 | mid | 40% | 100/100 | $12.46 |
| Mistral Large 2407 | mid | 26% | 100/100 | $25.74 |
| Arcee Trinity Large Thinking | mid | 31% | 100/100 | $105 |
| GPT-5.4 | frontier | 83% | 100/100 | $189 |
| Claude Opus 4.7 | frontier | 31% | 100/100 | $420 |
| Gemini 3.1 Pro Preview | frontier | 55% | 100/100 | $807 |
| Gemma 4 E4B IT | small | 23% | 100/100 | $1.04 |
| Granite 4.1 8B | small | 28% | 100/100 | $2.07 |
| Ministral 8B Instruct 2410 | small | 25% | 100/100 | $5.68 |
| Qwen3 4B Instruct 2507 | small | 31% | 100/100 | $9.98 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-08T22:06:03.517Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock