Blackbox vs Direct Azure
See how Azure upstream time and Blackbox Gateway overhead contribute to latency for gpt-6-astra and gpt-6.1-sol.
Latency recap
Analysis window: November 6, 2026, 14:26:00 PST through November 7, 2026, 01:18:27 PST.
All requests for both models in the available latency-profile logs were analyzed. The figures below are mean latencies for successful, non-aborted streaming requests, shown in milliseconds to three decimal places. Errors and non-streaming requests are excluded from the latency averages.
| Mean latency | Astra: Azure | Astra: Blackbox | Sol: Azure | Sol: Blackbox |
|---|---|---|---|---|
| First byte | 3,243.707 ms (91.4%) | 304.958 ms (8.6%) | 2,035.569 ms (87.2%) | 298.759 ms (12.8%) |
| Total duration | 12,521.489 ms (97.6%) | 304.958 ms (2.4%) | 36,715.445 ms (99.2%) | 298.759 ms (0.8%) |
Time to first byte
The bars split mean time to first byte into Azure upstream time and Blackbox Gateway overhead. Bar lengths use the same scale for both models.
gpt-6-astra
3,548.665 ms
gpt-6.1-sol
2,334.328 ms
Total duration
Mean time from the start of a request until it finishes, including the streamed response. Blackbox Gateway overhead accounts for 2.4% of Astra's total duration and 0.8% of Sol's.
gpt-6-astra
12,826.447 ms
gpt-6.1-sol
37,014.204 ms