SEC Risk-Factor Change Detection
Given prior-year and current-year risk-factor sections, identify added, removed, and materially changed risks with paired supporting passages.
View token rates · observed 9/11/2026
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 | 61% | 100/100 | $43.79 |
| Qwen3 235B A22B | mid | 71% | 100/100 | $368 |
| DeepSeek V3 | mid | 89% | 100/100 | $127 |
| Mistral Large 2407 | mid | 88% | 100/100 | $225 |
| Arcee Trinity Large Thinking | mid | 81% | 100/100 | $140 |
| GPT-5.4 | frontier | 65% | 100/100 | $3129 |
| Claude Opus 4.7 | frontier | 43% | 100/100 | $5548 |
| Gemini 3.1 Pro Preview | frontier | 7% | 100/100 | $2519 |
| Gemma 4 E4B IT | small | 78% | 100/100 | $14.83 |
| Granite 4.1 8B | small | 62% | 100/100 | $15.38 |
| Ministral 8B Instruct 2410 | small | 44% | 100/100 | $26.00 |
| Qwen3 4B Instruct 2507 | small | 38% | 100/100 | $127 |
LLM-judge pass rate on 100 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-08T21:19:02.986Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock