OpenAI Adds Cheaper GPT-6 Sol and Luna Models
The company is cutting API costs as it brings its latest model family to coding, agent workflows and high-volume tasks.
Topics
News
- Apple Tightens Mac Access Rules as AI Agents Gain Reach
- Trump Puts Intelligence Chief Jay Clayton in Charge of AI Task Force
- IBM Widens AI, Quantum Research With IIT Bombay, IISc
- OpenAI, Synopsys Build AI Model That Could Slash Chip Design Time
- Pentagon Plans Four-Star Command for AI, Drone Warfare
- California Draws the Line on AI Firings, Layoffs and Workplace Surveillance
OpenAI has launched GPT-6 Sol and GPT-6 Luna, adding faster and cheaper options to the GPT-6 family weeks after introducing its higher-end Astra model.
Sol is designed for complex coding and agentic workflows, while Luna is aimed at faster, high-volume tasks. Both are available through OpenAI’s API and are rolling out to ChatGPT Work and Codex. Free and Go users can access Luna through the desktop app.
The new models are substantially cheaper than their GPT-5.6 predecessors. Sol costs $2 per million input tokens and $10 per million output tokens, compared with promotional rates of $4 and $20 for GPT-5.6 Sol. Luna costs $0.10 for input and $0.50 for output, down from $0.20 and $1.20 for GPT-5.6 Luna.
“We trained GPT-6 Sol and Luna with similar methods as GPT-6 Astra,” OpenAI said, bringing some of Astra’s improvements in professional work, coding, computer use and alignment to lower-cost models.
In OpenAI’s tests, Sol scored 33.2% on AutomationBench at its xhigh reasoning setting, compared with 26.9% for Anthropic’s Claude Opus 5. On the DeepSWE v1.1 coding benchmark, Sol scored 68.8% and Luna 66.6% at their highest reported settings.
OpenAI also said Sol made about half as many factual errors as GPT-5.6 Sol in an internal test based on conversations in which users had previously flagged mistakes. The evaluation deliberately focused on difficult, error-prone cases and does not represent typical production traffic.
Both models support a context window of about 1.05 million tokens and reasoning settings ranging from none to max. OpenAI has also lowered the price of cached input to one-tenth of the standard input rate, at $0.20 per million tokens for Sol and $0.01 for Luna.
OpenAI is also improving prompt caching for the GPT-6 family, offering discounts of up to 90% on cached input-token reads for applications that repeatedly reuse the same context.
Astra remains OpenAI’s highest-end GPT-6 model, while Sol and Luna give developers cheaper options for workloads where cost and scale matter more. OpenAI describes Luna as its most efficient model for focused, high-volume tasks and Sol as a model built for complex coding and agentic workflows.
