OpenAI Launches GPT-6 Sol and GPT-6 Luna With Lower Prices, Improved Performance

OpenAI has launched GPT-6 Sol and GPT-6 Luna, two new models that extend some of the capabilities of its GPT-6 Astra model to lower-cost tiers. The models are aimed at professional work, coding, computer use and other tasks that require sustained AI assistance. OpenAI has also cut their API prices by 50 percent compared with the promotional pricing of their GPT-5.6 predecessors. GPT-6 Sol and Luna are available across Work, Codex and the API, while Luna is also available to Free and Go users through the desktop app.

OpenAI said GPT-6 Sol and GPT-6 Luna use training methods similar to those behind GPT-6 Astra and improve performance across professional work, factuality, coding, computer use and alignment. GPT-6 Astra remains the company's highest-capability model.

GPT-6 Sol and GPT-6 Luna Availability

GPT-6 Sol and Luna are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. Free and Go users can access GPT-6 Luna through the desktop app. The models are not yet available in standard ChatGPT, although OpenAI plans to roll them out gradually. They are also available through the API as gpt-6-sol and gpt-6-luna.

GPT-6 Sol costs $2 (roughly Rs. 191) per 1 million input tokens and $10 (roughly Rs. 957) per 1 million output tokens through the API. GPT-6 Luna is priced at $0.10 (roughly Rs. 10) per 1 million input tokens and $0.50 (roughly Rs. 48) per 1 million output tokens. OpenAI says both models are 50 percent cheaper than the promotional prices of GPT-5.6 Sol and GPT-5.6 Luna.

GPT-6 Sol and Luna Bring Performance and Cost Improvements

OpenAI says the new models improve performance across professional work, coding, computer use and factuality. On AutomationBench, GPT-6 Sol scored 33.2 percent at xhigh effort, compared with 26.9 percent for Claude Opus 5 at maximum effort, while costing $0.27 (roughly Rs. 26) per task. GPT-6 Luna improved by 5.4 percentage points over GPT-5.6 Luna at high effort, while reducing its cost per task by 58 percent.

GPT-6 Sol scored 68.8 percent on DeepSWE at maximum effort, compared with 69.9 percent for Claude Fable 5, while costing about 80 percent less per task. On OSWorld 2.0, Sol scored 60.5 percent at xhigh effort versus 60.3 percent for Claude Opus 5 at medium effort, with about 80 percent lower cost per task.

OpenAI says GPT-6 Sol makes about half as many mistakes as GPT-5.6 Sol in its internal factuality test. GPT-6 Luna can match GPT-5.6 Sol at higher effort at about one hundredth of the cost. OpenAI also says the new models provide clearer responses, use less jargon and include fewer unnecessary details.