OpenAI Introduces GPT-6 Sol and Luna with lower API prices

OpenAI has announced GPT-6 Sol and GPT-6 Luna, two new additions to the GPT-6 family. Sol is designed for complex coding, heavy reasoning, agentic workflows, and demanding professional tasks, while Luna targets high-volume workloads where lower latency and cost are more important than maximum reasoning capability. Both models use training methods related to GPT-6 Astra, bringing improvements in factuality, coding, computer use, and alignment.

On AutomationBench, GPT-6 Sol with xhigh reasoning effort scored 33.2%, compared with 26.9% for Claude Opus 5, at a reported cost of about $0.27 per task, around 9% of Opus 5's cost. On DeepSWE 1.1, Sol scored 68.8% versus 69.9% for Claude Fable 5, while Luna reached 66.6% with reported per-task costs 93% below Opus 5 and 96% below Fable 5.

API pricing starts at $2 per million short-context input tokens and $10 per million output tokens for Sol, while Luna costs $0.10 and $0.50 respectively. Long-context pricing is $4 and $15 for Sol, and $0.20 and $0.75 for Luna. The models are rolling out to paid ChatGPT plans, with Luna also available to Free and Go users through the desktop app. Enterprise administrators must enable them, and neither model is currently available in standard Chat.