OpenAI has expanded its GPT-6 model family with the launch of GPT-6 Sol and GPT-6 Luna, two new systems designed to bring frontier intelligence to a wider range of everyday work. The announcement follows the earlier release of GPT-6 Astra and focuses on combining strong performance with significantly lower costs. The new models are available starting today across ChatGPT Work, Codex, and the OpenAI API.
New Models Built for Cost Efficiency
GPT-6 Sol and GPT-6 Luna were trained with methods similar to those behind GPT-6 Astra, transferring advances in professional work, factuality, coding, computer use, and alignment to more affordable options. OpenAI says the GPT-6 models lead across the cost-to-intelligence curve and are supported by improvements in caching and inference. API prices have been cut by 50 percent compared with GPT-5.6 promotional pricing, with Sol at two dollars per million input tokens and Luna at ten cents per million input tokens.
Professional Work Performance
On the AutomationBench evaluation for business workflows, GPT-6 Sol at xhigh effort outperforms Claude Opus 5 at maximum effort while costing only nine percent of Opus 5 per task. GPT-6 Luna at high effort improves on its predecessor by 5.4 percentage points at 58 percent lower cost per task. In the Agents Last Exam benchmark, GPT-6 Sol at maximum effort scores 56.4 percent, above Claude Opus 5 at 60 percent lower cost per task.
Coding Performance
OpenAI also reports growing internal demand for coding agents, with median daily token usage valued at more than six hundred dollars per researcher. On FrontierCode, GPT-6 Sol improves substantially over GPT-5.6 Sol and matches Claude Fable 5.1 at xhigh effort at much lower cost. On DeepSWE v1.1, GPT-6 Sol scores 68.8 percent at maximum effort, within 1.1 points of Claude Fable 5 while reducing cost per task by approximately 80 percent.
Factuality and Computer Use
Factuality evaluations based on real world conversations show GPT-6 Sol making about half as many mistakes as its predecessor and approaching Astra level reliability at much lower cost. GPT-6 Luna also improves substantially and can match GPT-5.6 Sol at higher effort at about one hundredth of its cost. In computer use, GPT-6 Sol at xhigh effort achieves a similar OSWorld 2.0 offline score to Claude Opus 5 at medium effort while costing about 80 percent less per task.
Alignment and Communication Style
GPT-6 Sol and Luna build on the alignment work introduced with GPT-6 Astra, including lower rates of misleading claims about coding work in OpenAI evaluations. The company notes these evaluations test challenging situations and do not measure typical failure rates. The new models also adopt Astra's improved communication style, with clearer answers, less jargon, and fewer low value details.
Caching and Developer Tools
OpenAI has improved prompt caching for GPT-6 to deliver higher cache hit rates by default, giving agents faster responses and a 90 percent discount on cached input token reads. Developers can monitor caching through a new dashboard and diagnostic tool, adjust reasoning effort and tool availability without losing cache, and set explicit breakpoints to control prefix reuse. GitHub reports these improvements have reduced the share of prompt tokens requiring fresh processing by more than 50 percent across billions of requests.
Availability Across Products
GPT-6 Sol and GPT-6 Luna are available in ChatGPT Work and Codex starting today for all Plus, Pro, Business, Enterprise, and Edu users. Free and Go users can access GPT-6 Luna in the desktop app, while API users can call the models as gpt-6-sol and gpt-6-luna. The models are not yet available in Chat and will roll out gradually throughout the day to keep service stable.
With GPT-6 Sol and GPT-6 Luna, OpenAI is extending the capabilities of its most advanced model family to tasks that require a balance of intelligence, speed, and cost efficiency. The combination of lower API pricing, better caching, and stronger alignment could make frontier AI more practical for everyday applications and large scale deployments. The release signals OpenAI's continued effort to broaden access to advanced AI without compromising on performance or reliability.