GPT-6 Luna (Fast)

GPT-6 Luna (Fast) is presented on a third-party comparison page as the September 22, 2026 successor to GPT-5.6 Luna, positioned as the high-volume lane of the GPT-6 family rather than a coding-focused sibling like Sol or Astra. The page frames it as the Luna-tier drop-in for workloads that previously pinned gpt-5.6-luna, with Free and Go desktop app trials supported even though no API free tier exists. It is described as multimodal, accepting text and image input while producing text only, matching the catalog’s input/output modality profile.

The same source distinguishes GPT-6 Luna (Fast) from the Standard mode by noting that Fast runs at roughly twice the Standard throughput, making it the throughput-oriented choice within the Luna tier. Reasoning effort is documented across the full none-through-max range with a medium default, suggesting a flexible reasoning dial rather than a fixed setting. Because the only supplied evidence is a third-party comparison rather than an official model card, training scale, parameter count, and benchmark standing remain unspecified, and practical fit should be inferred from the high-volume, latency-sensitive use cases implied by the Fast designation.