Extraction
Clinical Trial Eligibility Extraction
Given a protocol, extract inclusion and exclusion criteria into typed fields while preserving thresholds, timing windows, and conditional logic.
Healthcare & Life SciencesPer protocol
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
512 in · 749 out · 11.16s p50
$8.98 /100K
Current price · TrustedRouter
View token rates · observed 9/11/2026
TrustedRouter · $0.0211 input · $0.1055 output /1M
39%
Cheapest
Qwen3 235B A22B
235B (22B active) · Off-the-shelf
513 in · 927 out · 17.99s p50
$192 /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
63%
GPT-5.4
Hosted · Off-the-shelf
481 in · 1,263 out · 8.14s p50
$2015 /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
91%
Best accuracy
All 12 model results and methodology
Generated 100/100 examples
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 17% | 100/100 | $40.59 |
| Qwen3 235B A22B | mid | 63% | 100/100 | $192 |
| DeepSeek V3 | mid | 42% | 100/100 | $136 |
| Mistral Large 2407 | mid | 18% | 100/100 | $149 |
| Arcee Trinity Large Thinking | mid | 36% | 100/100 | $118 |
| GPT-5.4 | frontier | 91% | 100/100 | $2015 |
| Claude Opus 4.7 | frontier | 63% | 100/100 | $3873 |
| Gemini 3.1 Pro Preview | frontier | 6% | 100/100 | $2431 |
| Gemma 4 E4B IT | small | 39% | 100/100 | $8.98 |
| Granite 4.1 8B | small | 38% | 100/100 | $9.70 |
| Ministral 8B Instruct 2410 | small | 0% | 100/100 | $14.05 |
| Qwen3 4B Instruct 2507 | small | 30% | 100/100 | $45.38 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-10T08:50:02.785Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock