Input/output modalities and key model metadata from the catalog.
30-day activity plus recent runtime. Text-first models use token volume; other modalities fallback to request activity.
Only benchmarks with comparable results across every selected model are shown.
Switch benchmark type to compare percent and numerical families separately.
Per-1M normalized pricing from observed provider tiers. Blended total uses 90% input + 10% output.
All unique meters observed across the selected models.
| Meter | Qwen 3.5 397B A17B Best option |
|---|---|
| Input Text Tokens | $0.35 |
| Output Text Tokens | $1.75 |
| Cached Read Text Tokens | $0.07 |
| Input Video Tokens | $0.60 |
API provider availability and subscription plans.
