🔍 Read the full analysis: The Price Drop Of GPT‑6 Sol And Luna: No Impact On Benchmark Performance 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 a 50% price reduction for GPT‑6 Sol and Luna models, effective September 22, 2026. Despite lower costs, benchmark performance remains roughly the same, though some regressions in specific tasks are noted.
OpenAI has reduced the prices of its GPT‑6 Sol and GPT‑6 Luna models by approximately 50%, effective September 22, 2026, with no significant change in their benchmark scores. This move aims to make AI more accessible for a broader range of applications by lowering operational costs while maintaining performance levels, according to the company.
On September 22, 2026, OpenAI launched GPT‑6 Sol and Luna at half the previous prices of their GPT‑5.6 predecessors. The new models benefit from improvements in caching and inference efficiencies, allowing OpenAI to pass savings to users. Specifically, GPT‑6 Sol now costs $2.00 per 1 million input tokens and $10.00 per 1 million output tokens, while Luna is priced at $0.10 and $0.50 for input and output, respectively. These reductions are a direct result of technical enhancements that reduce per-task costs, with caching discounts of up to 90% on input reads.
Independent evaluations by Artificial Analysis, published concurrently, confirm that the cost per task has roughly halved, with minimal change in model scores. GPT‑6 Sol at maximum effort scores 48 on the Artificial Analysis Intelligence Index, well above the median of 25 for comparable models, while Luna scores 37 against a median of 12. Despite the cost savings, some regressions were observed in specific knowledge-work evaluations, with GPT‑6 models dropping in performance on certain economic and productivity benchmarks, attributed to changes in output presentation quality.
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
Implications for Cost-Effective AI Deployment
The price reductions for GPT‑6 Sol and Luna mark a significant shift toward more affordable AI solutions for businesses and developers. By maintaining comparable benchmark performance, these models enable broader adoption in applications such as customer service, research, and automation, where cost constraints previously limited usage. The move emphasizes that AI affordability can improve without sacrificing core capabilities, potentially accelerating innovation and integration across industries.
However, the observed regressions in some knowledge and productivity benchmarks suggest that lower-cost models may require careful testing before deployment in tasks demanding high-quality, detailed outputs. The trade-off between cost and presentation quality could influence workflows that rely on producing polished documents or comprehensive reports.
As an affiliate, we earn on qualifying purchases.
Background on Model Pricing and Performance
OpenAI introduced GPT‑6 Astra on September 22, 2026, as part of its ongoing effort to democratize access to advanced AI. The Astra family includes models designed to deliver high performance at various price points, with GPT‑6 Sol and Luna positioned as more affordable options. The models’ pricing was previously aligned with GPT‑5.6, but recent technical improvements in caching and inference have enabled OpenAI to cut costs significantly.
Independent evaluations by Artificial Analysis, also released on September 22, 2026, confirmed that despite the price cuts, the models’ benchmark scores remained stable overall, with some regressions in specific knowledge tasks. The models’ core capabilities—such as reasoning, coding, and hallucination rates—were closely examined, revealing that while costs per task fell sharply, some quality metrics experienced minor declines, especially in detailed presentation and knowledge accuracy.
Unconfirmed Effects on Real-World Applications
It is not yet clear how these price reductions will impact long-term deployment in large-scale, mission-critical applications, as some regressions in detailed output quality have been observed. Further testing is needed to determine if these models meet the standards required for high-stakes use cases.Next Steps for Adoption and Performance Monitoring
OpenAI is expected to continue monitoring the performance of GPT‑6 Sol and Luna in real-world scenarios, with potential updates or tuning based on user feedback. Industry users are advised to conduct their own testing, especially for workflows demanding high-quality outputs, before fully integrating these models. Additionally, OpenAI may expand its caching and inference optimizations to further reduce costs or improve quality in future releases.
Key Questions
Will the lower prices affect the models’ capabilities?
According to OpenAI, the core capabilities remain stable, with benchmark scores roughly unchanged, though some specific evaluation areas have seen minor regressions.
Are there any trade-offs with the price reduction?
Yes, some regressions in detailed presentation quality and knowledge accuracy have been noted, especially in tasks requiring comprehensive outputs.
How might this affect AI deployment costs for businesses?
The significant cost savings could enable broader adoption and more frequent use of AI in various workflows, reducing overall operational expenses.
Will OpenAI release further updates to improve performance?
OpenAI is likely to continue refining caching and inference techniques, which could further lower costs or enhance output quality in future models.
Should I test these models before deploying them in critical tasks?
Yes, especially for workflows requiring high-quality, detailed outputs, as some regressions in presentation and knowledge accuracy have been observed.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
