🔍 Read the full analysis: OpenAI’s Cost Cuts For GPT‑6 Sol And Luna: Benchmark Performance Unchanged on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
OpenAI announced significant cost reductions for GPT-6 Sol and Luna models, halving their prices compared to GPT‑5.6. Despite the lower costs, benchmark performance remains largely unchanged, emphasizing improved cost efficiency over capability enhancements.
OpenAI has announced a 50% reduction in prices for its GPT‑6 Sol and Luna models, effective immediately. Despite the lower costs, the models’ benchmark performance scores remain roughly at the same level as their GPT‑5.6 predecessors, highlighting a focus on cost efficiency rather than new capabilities. This development is significant for businesses and developers aiming to integrate large language models into operational workflows at a lower expense, without sacrificing quality.
On September 22, 2026, OpenAI released GPT‑6 Sol and Luna models at half the previous prices, citing improvements in caching and inference that enable cost savings. The prices now stand at $2.00 per 1 million tokens for input and $10.00 for output in Sol, and $0.10 per input and $0.50 per output for Luna. These reductions are achieved through technical efficiencies, including enhanced caching strategies that offer up to 90% discounts on cached input reads, passing savings directly to users.
Independent analysis from Artificial Analysis confirms that the cost per task has roughly halved, with GPT‑6 Sol costing about $1.06 per task at maximum effort—compared to $1.99 for GPT‑5.6 Sol—and Luna costing approximately $0.07 per task, down from $0.20. Benchmark scores for these models remain strong: Sol scores 48 on the Artificial Analysis Intelligence Index, well above the median of 25 for models in its class, while Luna scores 37 against a median of 12. Performance in coding tasks has also improved slightly, with Sol scoring 57 on the Coding Agent Index, and Luna scoring 41, indicating maintained or improved capabilities despite the cost cuts.
However, some performance regressions are noted in specific knowledge-work evaluations, such as drops in Elo points on certain economic and multi-week benchmarks. These regressions are attributed to a reduction in presentation quality and detail in outputs, which may impact workflows requiring comprehensive deliverables. OpenAI’s own release notes mention a focus on shorter, more concise answers, which could influence model suitability depending on application needs.
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 AI Deployment Costs
The significant price reductions for GPT‑6 Sol and Luna expand options for deploying large language models across a broader range of applications, especially where cost was previously a limiting factor. For companies integrating AI into customer service, research, or automation workflows, these models offer a more affordable alternative without sacrificing benchmark performance, potentially enabling more extensive or more frequent AI usage. This shift could accelerate AI adoption in sectors sensitive to operational costs, such as small and medium enterprises or startups.
Moreover, the focus on cost efficiency over capability suggests that OpenAI aims to make AI tools more accessible, promoting widespread adoption and innovation. However, the noted regressions in some knowledge-based tasks highlight that cost savings may come with trade-offs in output detail and presentation quality, which organizations must consider when choosing models for specific workflows.
As an affiliate, we earn on qualifying purchases.
Background on GPT Model Pricing and Performance
OpenAI’s GPT models have historically been priced according to their capabilities, with newer versions typically offering improved performance at higher costs. The release of GPT‑6 Astra earlier this month set a high-performance benchmark, but also emphasized the importance of cost-effective models for practical deployment. Prior to this, GPT‑5.6 models represented a balance of capability and expense, with pricing that limited widespread use in cost-sensitive contexts.
The recent price cuts for GPT‑6 Sol and Luna mark a departure from the trend of equating higher performance with higher costs. Instead, OpenAI’s focus appears to be on making models more affordable through technical efficiencies, such as caching improvements and inference optimizations. This approach aligns with industry trends toward democratizing AI access, enabling a wider range of users to leverage advanced language models.
large language model API pricing plans
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Model Performance and Use Cases
It is not yet clear how these cost reductions will affect long-term model performance in complex, real-world applications that demand detailed, nuanced outputs. The observed regressions in knowledge-work benchmarks suggest potential limitations for workflows requiring comprehensive deliverables. Additionally, the impact of these models on broader AI market dynamics and competitive positioning remains to be seen, as other providers may respond with their own pricing strategies.
As an affiliate, we earn on qualifying purchases.
Next Steps for OpenAI and Model Adoption
OpenAI is expected to continue refining its caching and inference techniques to further reduce costs. Organizations interested in adopting GPT‑6 Sol and Luna should evaluate their specific workflow needs, especially regarding output detail and presentation quality. Monitoring user feedback and performance data over the coming months will be crucial to understanding how these models perform in diverse real-world scenarios. Additionally, OpenAI may release further updates or new models that balance cost and capability more effectively.
As an affiliate, we earn on qualifying purchases.
Key Questions
Will the performance of GPT‑6 Sol and Luna remain stable over time?
While initial evaluations show stable benchmark scores, ongoing performance in real-world applications will depend on further updates and how models are fine-tuned for specific tasks. Continuous monitoring is recommended.
Are there trade-offs in output quality with these cost reductions?
Yes, some regressions in detailed knowledge and presentation quality have been observed, particularly in complex tasks requiring comprehensive outputs. Shorter, more concise answers are emphasized.
How might this pricing strategy influence the AI market?
Lower prices for powerful models could increase adoption, especially among smaller organizations, and may prompt competitors to adjust their pricing strategies to stay competitive.
Will OpenAI release newer models that combine low cost with high capability?
OpenAI is likely to continue optimizing models for both cost and performance, but specific future releases have not been officially announced yet.
What should organizations consider before switching to these models?
Organizations should evaluate their need for detailed outputs versus shorter responses, and test the models’ performance in their specific workflows to ensure suitability.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
