Analysis
Three-Way Match Discrepancy Resolution
Given a purchase order, goods receipt, and invoice, identify quantity, price, tax, and vendor mismatches and recommend the permitted resolution code.
Finance & AccountingPer invoice
Not benchmarked
No result on this tier
Token profile pending
— /100K
Run this tier to add it
—
Gemma 4 E4B IT
E4B · Off-the-shelf
862 in · 992 out · 16.08s p50
$12.28 /100K
Current price · TrustedRouter
View token rates · observed 9/11/2026
TrustedRouter · $0.0211 input · $0.1055 output /1M
81%
Cheapest
DeepSeek V3
Hosted · Off-the-shelf
792 in · 814 out · 6.59s p50
$103 /100K
Best current price · TrustedRouter
Compare 2 router rates · observed 9/11/2026
TrustedRouter · $0.3376 input · $0.9390 output /1M
OpenRouter · $0.2900 input · $1.14 output /1M
85%
Claude Opus 4.7
Hosted · Off-the-shelf
1,265 in · 1,515 out · 16.15s p50
$4420 /100K
Best current price · OpenRouter
Compare 2 router rates · observed 9/11/2026
OpenRouter · $5.00 input · $25.00 output /1M
TrustedRouter · $5.28 input · $26.38 output /1M
93%
Best accuracy
All 12 model results and methodology
Generated 100/100 examples
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 68% | 100/100 | $36.15 |
| Qwen3 235B A22B | mid | 76% | 100/100 | $288 |
| DeepSeek V3 | mid | 85% | 100/100 | $103 |
| Mistral Large 2407 | mid | 77% | 100/100 | $196 |
| Arcee Trinity Large Thinking | mid | 41% | 100/100 | $160 |
| GPT-5.4 | frontier | 88% | 100/100 | $1686 |
| Claude Opus 4.7 | frontier | 93% | 100/100 | $4420 |
| Gemini 3.1 Pro Preview | frontier | 68% | 100/100 | $2142 |
| Gemma 4 E4B IT | small | 81% | 100/100 | $12.28 |
| Granite 4.1 8B | small | 60% | 100/100 | $12.91 |
| Ministral 8B Instruct 2410 | small | 53% | 100/100 | $27.52 |
| Qwen3 4B Instruct 2507 | small | 64% | 100/100 | $108 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-08T23:34:03.000Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock