🔍 Read the full analysis: OpenAI’s Cost Reduction For GPT‑6 Sol And Luna: What’s Staying The Same? on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI announced GPT‑6 Sol and Luna on September 22, 2026, offering models at 50% lower prices than GPT‑5.6. While costs have decreased significantly with comparable performance in some evaluations, certain knowledge tasks have seen regressions, and quality trade-offs are noted.
OpenAI has introduced GPT‑6 Sol and Luna models on September 22, 2026, with prices halved compared to GPT‑5.6. This move aims to make advanced AI more accessible by reducing costs while maintaining performance, marking a shift in how AI models are deployed in business workflows.
The new models, GPT‑6 Sol and Luna, are priced at 50% less than their GPT‑5.6 predecessors, with GPT‑6 Sol costing $2.00 per 1 million input tokens and $10.00 per 1 million output tokens, and GPT‑6 Luna costing $0.10 and $0.50 respectively. OpenAI attributes these savings to improvements in caching and inference techniques, which allow for lower operational costs. Despite the lower prices, independent analysis from Artificial Analysis indicates that the models’ performance in some AI benchmarks remains high, with GPT‑6 Sol scoring 48 on the Artificial Analysis Intelligence Index and Luna scoring 37, both well above median scores for models in their price classes. Cost per task has approximately halved, with GPT‑6 Sol at $1.06 and Luna at $0.07, compared to their GPT‑5.6 equivalents, even though both models now generate slightly more output tokens per task. However, some regressions in knowledge-based evaluations have been observed, with GPT‑6 Sol and Luna scoring lower on tasks measuring economic and work-related knowledge, which OpenAI attributes to changes in output presentation quality. Notably, hallucination rates have decreased significantly, with Sol reducing hallucinations from 92% to 60%, and Luna from 93% to 77%, primarily by declining to answer more questions, which also affects accuracy and completeness in some cases.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 of Cost Reduction on AI Deployment
This development significantly lowers barriers for businesses to incorporate advanced AI models into their workflows, enabling more tasks to be automated at a fraction of previous costs. The reduction in hallucinations and improvements in inference efficiency also suggest that models can be more reliable for customer-facing applications and research support, although some knowledge accuracy regressions warrant caution. Overall, this shift could accelerate AI adoption across sectors, making powerful models accessible to a broader range of users and use cases.
As an affiliate, we earn on qualifying purchases.
Historical Pricing and Performance Trends
OpenAI’s previous models, notably GPT‑5.6, set a high-performance standard but came with high operational costs, limiting widespread adoption for cost-sensitive applications. The release of Astra, the top-tier model, showcased the potential of larger, more capable models, but at significant expense. The move to introduce GPT‑6 Sol and Luna at half the price marks a strategic pivot toward democratization of AI, emphasizing cost efficiency alongside performance. Independent evaluations, such as those from Artificial Analysis, have consistently shown that while newer models improve in some metrics, they can also regress in others, especially in knowledge accuracy and presentation quality, reflecting ongoing challenges in balancing model size, cost, and quality.
As an affiliate, we earn on qualifying purchases.
Remaining Questions About Model Capabilities
It is still unclear how these models will perform in real-world, long-term deployments, especially in complex knowledge-based tasks or applications requiring high factual accuracy. The observed regressions in some benchmarks suggest potential limitations in certain domains, and the impact of reduced presentation quality on user experience remains to be fully understood. Additionally, the long-term effects of increased refusal rates on workflow efficiency are yet to be evaluated in operational settings.
AI model performance evaluation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Adoption and Evaluation
OpenAI is expected to continue refining GPT‑6 Sol and Luna based on user feedback and ongoing benchmarking. Organizations considering these models should conduct thorough testing, especially for knowledge-intensive tasks, before full deployment. Further updates may include improvements in factual accuracy, presentation quality, and cost efficiency, alongside expanded tooling for caching and inference management. Monitoring how these models perform in diverse real-world applications will determine their broader adoption in the coming months.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much cheaper are GPT‑6 Sol and Luna compared to GPT‑5.6?
GPT‑6 Sol and Luna are priced at approximately 50% less than GPT‑5.6, with GPT‑6 Sol costing $2.00 per 1 million input tokens and $10.00 per 1 million output tokens, and Luna costing $0.10 and $0.50 respectively.
What are the main performance differences between GPT‑6 Sol and Luna?
GPT‑6 Sol scores higher on intelligence benchmarks (48 vs. 37 on the AI Index) and is more suited for complex tasks, while Luna offers faster output (154 tokens/sec vs. 115) and is optimized for lower-cost, high-volume use cases. Both models show reduced hallucination rates and improved inference efficiency.
Are there any trade-offs with the new models?
Yes, some knowledge and presentation quality regressions have been observed, and GPT‑6 models tend to refuse answering more often, which can impact productivity in certain workflows. Users should evaluate these factors based on their specific needs.
Will the models’ reduced hallucinations improve reliability?
Reduced hallucination rates suggest improved factual reliability, especially for customer-facing and research applications, but users should still verify outputs in critical use cases due to remaining limitations and occasional regressions.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
