Independent inference-speed benchmarks for GPT-6 Luna, measured through the official OpenAI API: median output tokens per second, time to first token (TTFT), total response time and availability, refreshed every 15 minutes.
Insufficient data: only 2 successful sample(s) in this window. Figures may be unstable.
Region is the observed Cloudflare data center and is not a geographic guarantee. Measurements from different regions are never combined. See methodology.
ModelSpeedTest