OpenAI Unveils GPT-6 Sol and Luna: Half the Price, Same Frontier Intelligence

OpenAI Expands GPT-6 Family with Cost-Efficient Sol and Luna

OpenAI has officially expanded its GPT-6 lineup with the release of GPT-6 Sol and GPT-6 Luna, two new models designed to deliver frontier intelligence at half the cost of their predecessors. The announcement, made on September 22, 2026, follows the earlier launch of GPT-6 Astra, which remains the company's flagship for the most demanding tasks.

The new models represent a strategic shift toward cost efficiency, targeting developers and enterprises that need advanced AI capabilities without the premium price tag. By leveraging improvements in caching and inference, OpenAI has reduced API prices by 50% compared to GPT-5.6 promotional pricing, making these models accessible for high-volume and everyday applications.

Pricing Breakdown: A 50% Cut Across the Board

The headline change is the dramatic price reduction. GPT-6 Sol now costs $2 per million input tokens and $10 per million output tokens, down from $4 and $20 respectively for GPT-5.6 Sol. GPT-6 Luna follows suit, priced at $0.10 per million input tokens and $0.50 per million output tokens, halving the previous rates of $0.20 and $1.20.

These are not promotional prices but the standard rates, an OpenAI spokesperson confirmed to The New Stack. The reductions are made possible by enhanced prompt caching, which now offers 90% discounts on cached input-token reads, and more efficient inference infrastructure. GitHub has already reported that these improvements reduced fresh prompt processing by over 50% across billions of requests.

Performance Benchmarks: Near-Astra Quality at Fraction of Cost

OpenAI is positioning Sol and Luna as capable alternatives to Astra for most professional workflows. On AutomationBench, GPT-6 Sol at xhigh effort scored 33.2%—outperforming Claude Opus 5 at max effort (26.9%) while costing just 9% of Opus 5's price per task. Even at low effort, GPT-6 Astra scored 30.3% but at 3.9x Sol's cost.

In coding, DeepSWE v1.1 results show GPT-6 Sol at max effort reaching 68.8%, within 1.1 points of Claude Fable 5's highest score (69.9%) but at 80% lower cost per task. GPT-6 Luna also impressed, matching Claude Opus 5 and Fable 5 at medium effort while costing 93-96% less per task.

Factuality and Alignment: Fewer Mistakes, Better Collaboration

OpenAI emphasizes significant improvements in factual reliability. On internal evaluations based on real-world flagged conversations, GPT-6 Sol makes about half as many mistakes as its predecessor, approaching Astra-level accuracy. GPT-6 Luna also improves substantially, matching GPT-5.6 Sol's accuracy at higher effort levels but at roughly one-hundredth the cost.

The models also inherit Astra's refined collaboration style, offering clearer, more concise responses with less jargon. In a side-by-side comparison, GPT-6 Sol demonstrated better judgment by not jumping to conclusions and providing more transparent explanations of its actions—a notable upgrade for technical conversations.

Caching Improvements and Developer Tools

Beyond pricing, OpenAI has enhanced prompt caching to help developers save more. The new Prompt Caching Dashboard and diagnostics tool allow developers to monitor cache hit rates and identify optimization opportunities. Additionally, developers can now adjust reasoning effort and tool availability without breaking cache, preserving context for reuse.

These improvements are particularly valuable for agentic workflows, where long conversations and repeated context are common. The combination of lower token prices and better caching could reduce total costs by more than 50% for many applications, making GPT-6 models viable for tasks previously deemed too expensive.

Availability and Ecosystem Integration

GPT-6 Sol and Luna are now available in ChatGPT Work and Codex for all Plus, Pro, Business, Enterprise, and Edu users. Free and Go users can access GPT-6 Luna in the desktop app. In the OpenAI API, they are accessible as gpt-6-sol and gpt-6-luna.

GitHub Copilot has also integrated the new models, with GPT-6 Sol available to Pro+, Max, Business, and Enterprise plans, while GPT-6 Luna is accessible across all Copilot tiers. This broad availability signals OpenAI's commitment to embedding its models into everyday developer workflows.

The Competitive Landscape and Why It Matters

OpenAI's aggressive pricing directly challenges Anthropic's Claude models, which have been competing on performance and cost. By halving prices while maintaining near-flagship performance, OpenAI is making a clear play for the mass market of AI applications—from automation to coding assistants.

The launch also addresses growing concerns about AI cost and sustainability. As coding agents and automated workflows consume more tokens, the total cost of ownership becomes critical. GPT-6 Sol and Luna offer a pragmatic solution for teams that need reliability without the premium of Astra.

While Astra remains the best choice for the most demanding projects, Sol and Luna represent the democratization of frontier intelligence. For developers, this means more room to iterate, more ambitious projects, and ultimately, more value from AI investments.