🔍 Read the full analysis: Half-Price GPT‑6 Sol And Luna: OpenAI Keeps Benchmark Scores Steady on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
OpenAI has released GPT‑6 Sol and Luna models at 50% lower prices than GPT‑5.6, maintaining benchmark scores. This shift significantly reduces AI costs for businesses, though some quality regressions are noted.
OpenAI has introduced GPT‑6 Sol and Luna models at half the price of their GPT‑5.6 predecessors, with no loss in benchmark scores, marking a significant shift in AI cost efficiency. The models, launched on September 22, 2026, aim to make advanced AI more accessible for a broader range of applications, from customer service to research workflows.
Both GPT‑6 Sol and Luna are priced at approximately 50% less than GPT‑5.6 models, with GPT‑6 Sol costing $2.00 per 1 million input tokens and $10.00 per 1 million output tokens, and GPT‑6 Luna at $0.10 and $0.50 respectively. These reductions are achieved through improvements in caching and inference techniques, which enable OpenAI to lower operational costs while passing savings to users.
Independent analysis by Artificial Analysis confirms that the models’ performance remains roughly on par with previous versions, with GPT‑6 Sol scoring 48 on the Artificial Analysis Intelligence Index and Luna scoring 37, both well above their respective medians. Cost per task also drops significantly, with GPT‑6 Sol at approximately $1.06 and Luna at about $0.07, representing roughly 50-60% savings compared to GPT‑5.6.
However, some regressions were noted in knowledge-work evaluations, with GPT‑6 models scoring lower on certain economic and productivity benchmarks. Artificial Analysis attributes these declines to reduced presentation quality and shorter outputs, which may impact workflows requiring detailed deliverables. Despite these issues, improvements in hallucination reduction and refusal rates are notable, with Sol decreasing hallucinations from 92% to 60%, and Luna from 93% to 77%.
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 Cost and Deployment Strategies
The release of GPT‑6 Sol and Luna at half the previous cost dramatically lowers barriers for integrating advanced AI into products and workflows. For businesses, this means the potential to automate more tasks, reduce operational expenses, and expand AI-driven services without sacrificing performance. The unchanged benchmark scores demonstrate that these cost savings do not come at the expense of intelligence, making these models highly attractive for a wide range of applications.
This development could accelerate AI adoption across industries, especially in sectors where budget constraints previously limited AI use. However, the noted regressions in some knowledge and presentation tasks suggest that users should carefully evaluate these models for specific workflows, particularly those requiring detailed, well-structured outputs.
As an affiliate, we earn on qualifying purchases.
Background on Model Pricing and Performance
OpenAI’s previous models, including GPT‑5.6, set a high standard for AI performance but at a significant cost, limiting accessibility for smaller organizations. The company has emphasized that recent improvements in caching and inference efficiency enable these new models to operate at lower prices while maintaining comparable benchmark scores. The launch of Astra, the top-tier model, continues to represent the high-end of performance, but the real shift lies in the cost-effective middle tier with Sol and Luna.
Independent evaluations, such as those by Artificial Analysis, have shown that while the models’ scores remain robust, some quality aspects, especially in detailed knowledge tasks, have experienced slight regressions. This follows OpenAI’s own notes indicating a focus on reducing low-value details and answer length, which may explain some of the performance dips in certain benchmarks.
Unresolved Aspects of Model Performance and Adoption
It is not yet clear how these models will perform across diverse real-world tasks outside controlled benchmarks, especially in complex or nuanced workflows. The noted regressions in some knowledge and presentation benchmarks suggest that certain use cases may require further testing. Additionally, the long-term impact on the AI market and competitive landscape remains uncertain, as other providers may respond with their own cost reductions or performance improvements.
Next Steps for Users and OpenAI’s Strategy
Users should evaluate these models within their specific workflows, paying close attention to output quality and suitability. OpenAI is likely to continue refining caching and inference techniques, potentially further reducing costs or improving performance. Future updates may also address the current regressions, especially in detailed knowledge tasks. Monitoring adoption trends and feedback will be key to understanding how these models reshape AI deployment in the coming months.
Key Questions
How do GPT‑6 Sol and Luna compare to previous models in performance?
They maintain benchmark scores similar to GPT‑5.6, with Sol scoring 48 and Luna 37 on the Artificial Analysis Intelligence Index, indicating comparable intelligence levels.
What are the main benefits of the price reduction?
The models now cost about half as much per task, enabling broader deployment, cost savings, and more automation opportunities for businesses.
Are there any downsides to the new models?
Some evaluations show regressions in knowledge-work benchmarks and shorter, less detailed outputs, which may affect workflows requiring comprehensive deliverables.
Will these models replace higher-end options like Astra?
No, Astra remains the top-tier model for tasks demanding the highest quality, but Sol and Luna provide cost-effective alternatives for less demanding applications.
What should users do before switching to these models?
Test the models within their specific workflows to ensure output quality meets their needs, especially for tasks requiring detailed, well-structured responses.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
