🔍 Read the full analysis: Half-Price GPT‑6 Sol And Luna: OpenAI Maintains Benchmark Performance on ThorstenMeyerAI.com
Get smart everyday buys delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
TL;DR
OpenAI has introduced GPT‑6 Sol and Luna models at 50% lower prices than GPT‑5.6. These models match previous performance levels while significantly reducing costs, potentially transforming AI application economics.
OpenAI has launched GPT‑6 Sol and Luna on September 22, 2026, priced at half the cost of their GPT‑5.6 predecessors, while maintaining comparable benchmark performance. This development marks a significant shift in AI economics, making advanced models more accessible for a broader range of applications.
The new models, GPT‑6 Sol and Luna, are designed to democratize AI deployment by drastically reducing costs. GPT‑6 Sol costs $2.00 per 1 million tokens for input and $10.00 for output, while Luna costs $0.10 and $0.50 respectively, representing approximately 50-60% savings compared to GPT‑5.6. OpenAI attributes these savings to improvements in caching and inference technologies, which lower operational expenses.
Independent analysis by Artificial Analysis confirms that the cost per task has halved, with GPT‑6 Sol at $1.06 and Luna at $0.07 per task at maximum effort, while performance scores on various benchmarks remain high. For example, Sol scores 48 on the Artificial Analysis Intelligence Index, significantly above the median of 25 for models in its price class, with Luna scoring 37. On coding tasks, Sol scores 57 on the Coding Agent Index, showing gains over previous versions, though some regressions appear in knowledge-work evaluations.
Quality improvements are most evident in hallucination reduction, with Sol decreasing hallucination rates from 92% to 60%, and Luna from 93% to 77%. However, these models also tend to decline in delivering comprehensive, well-presented outputs, with some evaluations indicating regressions in presentation quality and completeness. OpenAI reports that Sol makes about half as many factual errors as its predecessor, but also attempts fewer questions, which can impact use cases requiring detailed responses.
GPT‑6 Sol and Luna: half the price, about the same intelligence
OpenAI’s September 22, 2026 release doesn’t raise the ceiling. It lowers the cost of everything below it, which changes what’s worth automating.
Per 1M input / output tokens. Cached input reads keep the 90% discount.
Cost per task, halved
Measured by Artificial Analysis as the weighted cost of one Intelligence Index task, at max effort.
The effort dial moves cost more than the model choice
| Model and effort | Intelligence Index | Cost per task |
|---|---|---|
| GPT‑6 Sol (max) | 48 | $1.06 |
| GPT‑6 Sol (low) | 34 | $0.13 |
| GPT‑6 Luna (max) | 37 | $0.07 |
| GPT‑6 Luna (low) | 21 | $0.0045 |
| GPT‑6 Luna (non‑reasoning) | 18 | $0.01 |
Sol at low effort keeps about 70% of its max score for roughly an eighth of the cost, because it writes far fewer reasoning tokens. For reference, Claude Opus 5.5 leads the same index at 58.
What got better, and what got worse
Better
- Hallucination rate on AA‑Omniscience: Sol 92% → 60%, Luna 93% → 77%
- Coding Agent Index: Sol 57, up 2 points, at ~50% lower cost per task
- OpenAI reports about half as many factual mistakes for Sol as its predecessor
- Higher cache hit rates; GitHub reports over 50% fewer prompt tokens needing fresh processing
Sol gets there partly by declining more: it attempts 83% of questions vs 99%, and accuracy falls 59% → 54%.
Worse
- GDPval‑AA v2.1: Sol down ~100 Elo, Luna down ~75
- AA‑Briefcase v1.1: Luna down ~45 Elo
- Coding Agent Index: Luna 41, down 2 points
- Both models write more output tokens per task than their predecessors
Reviewers attribute the drops to weaker presentation and deliverables that omit required elements.
What to do about it
Impact on AI Deployment and Cost Efficiency
The release of GPT‑6 Sol and Luna at half the previous prices significantly lowers barriers to deploying advanced AI models across industries. This shift enables smaller companies and projects with limited budgets to incorporate AI at scale, potentially accelerating AI adoption in sectors like customer service, research, and automation. The models’ maintained performance levels mean users can leverage cost savings without sacrificing quality, making AI more viable for tasks previously deemed too expensive.
Moreover, the improved caching and inference techniques demonstrate how technological advances can reduce operational costs, setting a new standard for AI model efficiency. This could lead to a competitive landscape where cost becomes a primary differentiator, prompting other providers to follow suit or innovate further to stay competitive.
As an affiliate, we earn on qualifying purchases.
Background on AI Model Cost and Performance Trends
Prior to this release, AI models like OpenAI’s GPT‑5.6 and Anthropic’s Claude Opus 5.5 set benchmarks for performance and cost. While larger, more capable models typically commanded higher prices, recent trends have emphasized balancing cost with quality. OpenAI’s Astra model, introduced earlier, set a high-performance standard, but the new Sol and Luna models shift focus toward affordability without significant performance trade-offs.
Historically, AI model costs have been driven by hardware expenses, inference efficiency, and caching strategies. The recent improvements in caching, as highlighted by OpenAI, demonstrate how optimization at the infrastructure level can substantially reduce costs, making high-performance AI accessible to a broader user base.
Independent evaluations confirm that while the models maintain strong benchmark scores, some areas like knowledge accuracy and presentation quality have experienced slight regressions, reflecting the ongoing trade-offs in model tuning and deployment strategies.
affordable AI chatbot development tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Aspects of Model Performance and Adoption
It is still unclear how these models will perform in large-scale, real-world deployments over extended periods. While benchmark scores are promising, actual use cases may reveal additional limitations, especially in complex or nuanced tasks. The long-term impact on AI ecosystem dynamics, including competitive responses and user adoption rates, remains to be seen.
Furthermore, the extent to which the models’ reduced presentation quality affects enterprise workflows requiring detailed, polished outputs is still under observation. OpenAI has indicated some regressions in this area, but the practical significance will depend on specific application needs.
As an affiliate, we earn on qualifying purchases.
Future Developments and Adoption Roadmap
OpenAI is expected to continue refining GPT‑6 Sol and Luna, addressing identified regressions and optimizing for specific use cases. Industry observers anticipate increased adoption in sectors that benefit from high-volume, cost-sensitive AI applications, such as customer support, content generation, and research automation.
OpenAI might also expand caching and inference improvements to other models, further lowering operational costs. Monitoring how competitors respond with their own cost-efficient models will be crucial, as will tracking real-world deployment outcomes to evaluate the models’ long-term viability and impact.
As an affiliate, we earn on qualifying purchases.
Key Questions
How do GPT‑6 Sol and Luna compare to previous models in performance?
Benchmark tests show that GPT‑6 Sol and Luna maintain high performance levels, with Sol scoring 48 and Luna 37 on the Artificial Analysis Index, comparable to or better than previous models, though some evaluations indicate minor regressions in knowledge tasks.
What are the main cost savings of these new models?
Cost per task is approximately 50-60% lower than GPT‑5.6, with GPT‑6 Sol costing about $1.06 per task and Luna about $0.07, thanks to improved caching and inference efficiencies.
Are there any trade-offs in quality or capability?
Yes, some evaluations show regressions in presentation quality and knowledge accuracy, and the models tend to attempt fewer questions, which could impact use cases requiring detailed, comprehensive outputs.
What does this mean for AI adoption in industry?
The lower costs make advanced AI models accessible to smaller organizations and broader applications, potentially accelerating adoption across sectors like customer service, research, and automation.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
