Corporate Minutes Public Redaction
Given synthetic board minutes and a redaction policy, remove personal data and confidential strategy while preserving motions, votes, and publishable governance facts.
View token rates
Compare 2 router rates · observed 9/11/2026
Compare 2 router rates · observed 9/11/2026
All 12 model results and methodology
Generated 100/100 examples
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 16% | 100/100 | $41.58 |
| Qwen3 235B A22B | mid | 41% | 100/100 | $131 |
| DeepSeek V3 | mid | 38% | 100/100 | $60.91 |
| Mistral Large 2407 | mid | 29% | 100/100 | $128 |
| Arcee Trinity Large Thinking | mid | 26% | 100/100 | $88.28 |
| GPT-5.4 | frontier | 62% | 100/100 | $782 |
| Claude Opus 4.7 | frontier | 64% | 100/100 | $2636 |
| Gemini 3.1 Pro Preview | frontier | 7% | 100/100 | $2533 |
| Gemma 4 E4B IT | small | 9% | 100/100 | $6.23 |
| Granite 4.1 8B | small | 10% | 100/100 | $8.86 |
| Ministral 8B Instruct 2410 | small | 5% | 100/100 | $19.00 |
| Qwen3 4B Instruct 2507 | small | 14% | 100/100 | $48.17 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-08T21:03:02.952Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock