Extraction
Invoice Schema Extraction
Extract caller-selected invoice fields into exactly typed JSON, preserving source values and returning null for absent fields. Available model: neurometric/document-structured-extraction.
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
469 in · 241 out · 4.25s p50
$3.53 /100K
Current price · TrustedRouter
View token rates · observed 9/11/2026
TrustedRouter · $0.0211 input · $0.1055 output /1M
60%
Cheapest
Mistral Large 2407
123B · Off-the-shelf
517 in · 260 out · 7.17s p50
$68.42 /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
63%
GPT-5.4
Hosted · Off-the-shelf
394 in · 179 out · 1.77s p50
$367 /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
73%
All 12 model results and methodology
Generated 100/100 examples
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 0% | 100/100 | $31.33 |
| Qwen3 235B A22B | mid | 49% | 100/100 | $58.83 |
| DeepSeek V3 | mid | 60% | 100/100 | $30.63 |
| Mistral Large 2407 | mid | 63% | 100/100 | $68.42 |
| Arcee Trinity Large Thinking | mid | 52% | 100/100 | $122 |
| GPT-5.4 | frontier | 73% | 100/100 | $367 |
| Claude Opus 4.7 | frontier | 67% | 100/100 | $987 |
| Gemini 3.1 Pro Preview | frontier | 28% | 100/100 | $1797 |
| Gemma 4 E4B IT | small | 60% | 100/100 | $3.53 |
| Granite 4.1 8B | small | 32% | 100/100 | $3.81 |
| Ministral 8B Instruct 2410 | small | 24% | 100/100 | $10.01 |
| Qwen3 4B Instruct 2507 | small | 18% | 100/100 | $19.79 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-08T23:06:02.658Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock