Model Tradeoffs
Recent models and measurement coverage
| Model | Input / 1M | Cached input / 1M | Output / 1M | Comparison status |
|---|---|---|---|---|
| Loading live catalog... | ||||
Standard provider rates are shown before BrightWrapper markup. A frontier point needs independent task-quality, time, and cost measurements for the same configuration. Automatic routing uses complete measured configurations only.
Model Tradeoff Frontier
Calculation notes
Only configurations with published Artificial Analysis Intelligence, task-time, and task-cost measurements are plotted. Cost is Artificial Analysis's weighted average cost per Intelligence Index task. The fixed 100k-input, 20k-output request estimate remains in each model's details.
Fixed-request examples use the listed token rates. DeepSeek V4 Pro uses its peak rate because its off-peak discount depends on request time. Artificial Analysis task costs come from benchmark snapshots and may differ from today's token rates.
Quality and task cost use the September 5 Artificial Analysis v4.2 snapshot where available. Other task-cost results and task-time results come from the August 6 snapshot.
The routing table shows defaults; request-supplied weights can change the selected model.
Text-model routing paths
| Call | Endpoint | Priority map | Current selection | Use |
|---|