Completed: **three fresh agents, identical 15 routing judgments, medium reasoning, one run each.** | Model | Elapsed |

Completed: **three fresh agents, identical 15 routing judgments, medium reasoning, one run each.** | Model | Elapsed | Input tokens | Cached input¹ | Output² | Estimated cost: standard / fast | |---|---:|---:|---:|---:|---:| | **Jev** | **0.267 s** | **15,086** | — | **233** | **$0.000634** | | Terra | 37.2 s | 198,306 | 140,544 | 1,637 | $0.163 / $0.327 | | Sol | 79.5 s | 146,323 | 87,936 | 3,132 | $0.331 / $0.663 | | Astra | 28.2 s | 138,473 | 82,432 | 490 | $0.667 / $1.335 | ¹ Already included in input; charged at discounted rates. ² Includes reasoning tokens. The current **standard prices per million tokens** are: | Model | Input | Cached input | Output | |---|---:|---:|---:| | Jev | $0.042 | — | $0 | | Terra | $2 | $0.20 | $12 | | Sol | $4 | $0.40 | $20 | | Astra | $10 | $1 | $50 | GPT Fast API rates are twice these standard rates. These are **API-equivalent estimates, not measured invoice charges**; the agents ran through