OpenAI has expanded its GPT-6 lineup with GPT-6 Sol and GPT-6 Luna, two new models designed for professional work, coding, automation and computer-use tasks at lower API prices than their GPT-5.6 predecessors.
The models are priced at $2 per million input tokens and $10 per million output tokens for GPT-6 Sol, while GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens. OpenAI says the new prices represent a 50% reduction from the promotional pricing of the corresponding GPT-5.6 models.
OpenAI attributes the lower costs to improvements in inference and prompt caching. The company says its upgraded caching system delivers higher cache-hit rates and gives developers more control over cached prompts, potentially reducing the amount of processing required for repeated or long-running AI workloads.
The two models extend capabilities introduced with GPT-6 Astra, which OpenAI describes as its highest-capability model. Sol is positioned for complex coding and agentic workflows, while Luna is designed for more focused, high-volume tasks where cost and efficiency are important. Both support text and image inputs and can be used through OpenAI’s API.
Performance and Coding
OpenAI says GPT-6 Sol also delivers improvements across professional workflows, coding and computer-use tasks. On AutomationBench, the company reported a 33.2% score for Sol at xhigh reasoning effort, with an estimated cost of $0.27 per task. OpenAI also reported a 56.4% result for Sol on Agents’ Last Exam, an evaluation covering long-running professional tasks across 55 industries.
For software development, OpenAI reported a 68.8% score for GPT-6 Sol on DeepSWE v1.1 at maximum effort. GPT-6 Luna scored 66.6% on the same evaluation. OpenAI’s comparisons with competing models are based on its own testing and cost calculations, so the results should be considered within the methodology of those evaluations.
The company also reported improvements in computer-use performance. On the offline OSWorld 2.0 evaluation, GPT-6 Sol scored 60.5% at xhigh effort, compared with 60.3% for Claude Opus 5 at medium effort, according to OpenAI.
Lower-Cost AI for Developers
OpenAI has also changed how developers can manage prompt caching with the GPT-6 family. Cached input is priced at $0.20 per million tokens for Sol and $0.01 for Luna, significantly below their standard input rates. Developers can monitor cache performance, diagnose cache misses and specify points where prompt prefixes should be cached.
OpenAI says these changes are particularly relevant for applications that repeatedly process large amounts of context, including AI agents and automated business workflows.
The company also reported improvements over GPT-5.6 in internal alignment evaluations involving scenarios such as coding deception, broken searches and attempts to bypass safeguards. OpenAI cautions that these tests are deliberately designed to create difficult situations and should not be interpreted as representing failure rates during ordinary use.
GPT-6 Sol and GPT-6 Luna are available through the API as gpt-6-sol and gpt-6-luna. They are also rolling out across ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users, while Free and Go users can access GPT-6 Luna through the desktop app.
With Sol and Luna, OpenAI is expanding the GPT-6 family beyond its highest-capability model and putting greater emphasis on the cost of running AI at scale. For developers building agents, coding tools and automated workflows, the lower API prices are a central part of the new models’ positioning.
