1 cheaper models approved across 1 team; 1 team is still learning.
Sample workspace: these settings and check results are illustrative. In a real organisation, a request sent as parezen/auto goes to the cheapest approved model, and every check runs on your own traffic.
| Model | Cheaper than usual | Status | |
|---|---|---|---|
| Gemini 2.5 Flash | 17% | 97.4% | Approved |
| Model | Cheaper than usual | Status | |
|---|---|---|---|
| Gemini 2.5 Flash | 93% | — | Being checked (12 of 50 checks) |
| Llama 3.3 70B | 91% | — | Being checked (21 of 50 checks) |
| DeepSeek V3 | 86% | — | Being checked (30 of 50 checks) |
