WHAT CHANGED
OpenAI introduced GPT-6 Sol and GPT-6 Luna. Sol is positioned for complex coding and agentic workflows at a lower unit cost than Astra, while Luna is positioned for focused, high-volume, and cost-sensitive workloads.
The two-model release creates a documented choice between a general-purpose tier for complex coding and agentic work and an efficiency-focused tier for cost-sensitive volume. Although they share the same stated context window, output limit, reasoning-effort settings, and input/output modalities, their standard token prices differ substantially. Teams can therefore evaluate workload fit and token economics separately from core context and modality requirements.
Under standard API pricing, GPT-6 Luna is priced at $0.10 per 1 million input tokens and $0.50 per 1 million output tokens, versus $2 and $10 for GPT-6 Sol. Both models apply long-context pricing when total input exceeds 272,000 tokens: input, cached input, and cache-write prices are doubled, while output pricing is multiplied by 1.5 for the entire request. This makes request-level input size relevant when estimating costs, even though both models list a 1.05 million-token context window.
API teams choosing between an OpenAI model for complex coding or agentic workflows and one for focused, high-volume, or cost-sensitive work should compare GPT-6 Sol and GPT-6 Luna. Teams handling large prompts should also review the documented long-context pricing rule, which applies once total input exceeds 272,000 tokens.