OpenAI Unveils GPT-6 Sol & Luna: Powering AI with 50% Cheaper API & Enhanced Performance

Introduction: Affordable AI Gets a Major Upgrade

At a glance, OpenAI is once again pushing the boundaries of accessible AI with the release of GPT-6 Sol and GPT-6 Luna. These two new models join the acclaimed GPT-6 family, sitting just below the flagship GPT-6 Astra. Designed to bring Astra’s advanced capabilities to a wider audience, Sol and Luna promise significant performance improvements coupled with dramatically reduced API pricing, making powerful AI more affordable and efficient than ever for developers and businesses.

Availability and Strategic Positioning

Meanwhile, Sol and Luna are immediately available through the OpenAI API as gpt-6-sol and gpt-6-luna. Notably these are API-only models, meaning they are served directly by OpenAI without the option for self-hosting their weights.

The GPT-6 family now offers a tiered approach, allowing users to select the optimal model for their specific needs:

• GPT-6 Astra: The premier model, designed for the most demanding and complex tasks.
• GPT-6 Sol: Optimized for intricate coding challenges and professional-grade applications, offering a balance of power and cost-efficiency.
• GPT-6 Luna: Tailored for high-volume, everyday operations where speed and affordability are paramount.

Revolutionary API Pricing: Halving the Cost of Advanced AI

In practical terms, OpenAI has implemented substantial price reductions, making these new models incredibly attractive. Thanks to advancements in caching and inference, Sol and Luna API prices have been slashed by approximately 50% compared to their GPT-5.6 promotional rates.

Here’s a breakdown of the new pricing per 1 million tokens:

• GPT-6 Astra: Input $10.00, Output $50.00
• GPT-6 Sol: Input $2.00 (down from $4.00), Output $10.00 (down from $20.00)
• GPT-6 Luna: Input $0.10 (down from $0.20), Output $0.50 (down from $1.20)

For example, Notably, Luna’s output price reduction is even more significant, dropping by about 58%, further enhancing its cost-effectiveness for high-volume tasks.

Unprecedented Performance Benchmarks Across Key Areas

Sol and Luna don’t just offer lower prices; they deliver impressive performance that often surpasses competitors at a fraction of the cost across various benchmarks.

Professional Work Tasks

• On AutomationBench 1.0.6, Sol achieved a 33.2% score at a cost of just $0.27 per task. This significantly outperforms Claude Opus 5, which scored 26.9% at 11.1 times Sol’s cost. Even a low-effort Astra scored 30.3% at 3.9 times Sol’s cost.
• For agentic tasks, Sol scored 56.4% on Agents’ Last Exam, besting Claude Opus 5’s top score at 60% lower cost per task.

Advanced Coding Capabilities

• In DeepSWE v1.1, Sol achieved 68.8%, coming in just 1.1 points behind Claude Fable 5 but at approximately 80% lower cost per task.
• Luna also showcased strong coding prowess, scoring 66.6% on DeepSWE v1.1. This performance is comparable to Opus 5 and Fable 5 (at medium effort) but at an astonishing 93% and 96% lower cost per task, respectively.
• On FrontierCode 1.1 Main, which assesses production-ready code, Sol matched Claude Fable 5.1’s performance at a substantially reduced cost.

Computer Use and Factuality

• For complex computer interaction tasks, Sol scored 60.5% on OSWorld 2.0 offline, slightly outperforming Opus 5 (60.3%) at about 80% less cost per task.
• Luna also impressed, beating GPT-5.6 Sol on OSWorld 2.0 at just one-tenth of the cost.
• In terms of factuality, an internal OpenAI test revealed Sol makes about half as many mistakes as its predecessor. Luna, at higher effort, matched GPT-5.6 Sol’s accuracy at merely 1/100th of its cost.

That said, Beyond raw performance, OpenAI has also ensured that Sol and Luna inherit Astra’s refined communication style, promising clearer, more concise responses with less jargon, particularly in technical and coding conversations.

Advanced Prompt Caching for Developers

Recognizing the needs of long-running AI agents that frequently resend similar instructions, OpenAI has significantly upgraded its prompt caching system. This enhancement leads to higher cache hit rates by default and offers discounts of up to 90% on cached input reads. Shared prefixes reused within a 30-minute window are now eligible for these savings.

Developers gain new granular controls to optimize caching:

• A dedicated Prompt Caching Dashboard to monitor hit rates over time.
• A diagnostics tool that explains cache misses (e.g., “reason”: “tools_changed”).
• Explicit breakpoints, allowing precise control over where a cached prefix ends.
• The ability to change reasoning effort mid-conversation via configuration_update without invalidating the cache.
• allowed_tools to restrict callable tools while maintaining stable definitions.
• Prewarming capabilities to prepare known context before the initial user request.

These improvements are already showing tangible benefits, with reports indicating a cut of over 50% in prompt tokens requiring fresh processing, leading to faster responses in applications like GitHub Copilot.

Widespread Availability Across OpenAI’s Ecosystem

However, OpenAI is rolling out GPT-6 Sol and Luna across its ecosystem, making them accessible to a broad user base:

• API Access: Immediately available via gpt-6-sol and gpt-6-luna.
• ChatGPT Work and Codex: Accessible to Plus, Pro, Business, Enterprise, and Edu users.
• ChatGPT Desktop App: Luna is available for Free and Go users.
• ChatGPT Web Interface: A gradual rollout is underway through launch day.

Conclusion: A New Era of Accessible and Powerful AI

OpenAI’s release of GPT-6 Sol and Luna marks a significant step towards making advanced AI more accessible and cost-effective. With impressive performance benchmarks, revolutionary pricing, and sophisticated developer tools like enhanced prompt caching, these models are set to empower a new wave of innovation across diverse applications. Developers and businesses can now leverage cutting-edge AI capabilities with unprecedented efficiency and affordability.

Expert Perspective

From an industry angle, the clearest signal around GPT-6 Sol Luna is how it may influence luna. The story reads less like a one-day spike and more like a marker of broader movement.

The next phase will depend on how quickly teams, regulators, or customers react. In practice, that gives GPT-6 Sol Luna room to reshape expectations across cost over the near term.

For readers focused on practical impact, the best next step is to watch what changes around openai once attention turns into execution.

Frequently Asked Questions

Why does GPT-6 Sol Luna matter right now?

Introduction: Affordable AI Gets a Major Upgrade At a glance, OpenAI is once again pushing the boundaries of accessible AI with the release of GPT-6 Sol and GPT-6 Luna.

What broader change could GPT-6 Sol Luna signal?

These two new models join the acclaimed GPT-6 family, sitting just below the flagship GPT-6 Astra.

What should the market watch next around GPT-6 Sol Luna?

Designed to bring Astra’s advanced capabilities to a wider audience, Sol and Luna promise significant performance improvements coupled with dramatically reduced API pricing, making powerful AI more affordable and efficient than ever for developers and businesses.

Source: https://www.marktechpost.com/2026/09/22/openai-releases-gpt-6-sol-and-luna-50-cheaper-api-pricing-and-benchmarks/