Meeting-Prep
Generates a pre-meeting briefing from CRM contact and company data: goals, risks, and talking points.
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 99/100 examples — partial set accepted at the success threshold
| Model | Tier | Quality | Judged | Scenario cost |
|---|---|---|---|---|
| Nemotron Nano 9B v2 | small | 8% | 99/99 | $33.51 |
| Qwen3 235B A22B | mid | 4% | 99/99 | $192 |
| DeepSeek V3 | mid | 31% | 99/99 | $84.01 |
| Mistral Large 2407 | mid | 54% | 99/99 | $153 |
| Arcee Trinity Large Thinking | mid | 8% | 99/99 | $102 |
| GPT-5.4 | frontier | 100% | 99/99 | $1460 |
| Claude Opus 4.7 | frontier | 83% | 99/99 | $4083 |
| Gemini 3.1 Pro Preview | frontier | 60% | 99/99 | $2003 |
| Gemma 4 E4B IT | small | 39% | 99/99 | $6.19 |
| Granite 4.1 8B | small | 17% | 99/99 | $11.01 |
| Ministral 8B Instruct 2410 | small | 64% | 99/99 | $18.11 |
| Qwen3 4B Instruct 2507 | small | 1% | 99/99 | $57.94 |
LLM-judge pass rate on 99 synthetic examples. Generator: gpt-5.2. Judge: gpt-5.2. Evaluated 2026-09-06T15:36:02.594Z.
Directional: measured on a synthetic eval set generated by drydock. Cost/latency are not yet captured for taskrouter-run benchmarks.
drydock