GPT‑6 Sol And Luna Now Half The Cost, Benchmark Performance Remains The Same
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: GPT‑6 Sol And Luna Now Half The Cost, Benchmark Performance Remains The Same on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

OpenAI has launched GPT‑6 Sol and Luna models at half the previous prices, with benchmark performance remaining stable. Cost reductions are driven by improved caching and inference, making AI more accessible without sacrificing quality.

OpenAI has introduced GPT‑6 Sol and Luna models at **half the previous cost**, with benchmark performance remaining steady, marking a significant shift in AI affordability. The models, announced on September 22, 2026, aim to expand AI deployment by reducing expenses for businesses and developers, without sacrificing the models’ capabilities. This development underscores a strategic focus on cost efficiency, making advanced AI more accessible for a broader range of applications.

OpenAI’s latest models, GPT‑6 Sol and Luna, are now priced at **50% less** than their GPT‑5.6 predecessors. GPT‑6 Sol costs $2.00 per 1 million tokens for input and $10.00 for output, down from $4 and $20, respectively. Luna costs $0.10 for input and $0.50 for output, halved from previous prices of $0.20 and $1.20. These reductions are attributed to improvements in caching and inference techniques, which enable the models to be served at lower costs while maintaining performance levels. Independent analysis by Artificial Analysis confirms that, despite the halved costs, the models’ benchmark scores remain comparable to prior versions, with Sol scoring 48 and Luna 37 on their respective intelligence indices, well above their price-class medians. The models also show improved hallucination rates, with Sol reducing hallucinations from 92% to 60%, and Luna from 93% to 77%, mainly by declining to answer some questions rather than answering inaccurately. However, some evaluations indicate regressions in knowledge work tasks, with models scoring lower on certain economic and productivity benchmarks, possibly due to changes in output presentation quality.

At a glance
updateWhen: announced September 22, 2026
The developmentOpenAI announced on September 22, 2026, that GPT‑6 Sol and Luna models are now available at 50% lower prices, with performance benchmarks unchanged.

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.

GPT‑6 Sol
$4 / $20 → $2 / $10
GPT‑6 Luna
$0.20 / $1.20 → $0.10 / $0.50

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.

GPT‑5.6 Sol
$1.99
GPT‑6 Sol
$1.06
GPT‑5.6 Luna
$0.18
GPT‑6 Luna
$0.07

The effort dial moves cost more than the model choice

Model and effortIntelligence IndexCost 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

Already on GPT‑5.6 Sol or Luna? The move is mostly a price cut. Re‑test first if your output is a document someone reads, not data a system consumes.
Shelved an automation on cost? Token prices halved and the effort dial adds another order of magnitude. Re‑run the business case.
Choosing between labs? The question is no longer which model is smartest, but which clears your quality bar at the lowest cost per task.
ThorstenMeyerAI.comSources: OpenAI (pricing, vendor benchmarks) and Artificial Analysis (independent evaluation and model pages). Figures as of 23 September 2026.

Implications for AI Deployment and Cost Efficiency

This price reduction significantly broadens the potential for AI integration across industries by lowering operational costs. For companies and developers, the models’ stable performance at half the cost means more tasks can be automated, and AI-driven workflows become more economically viable. The shift emphasizes that cost efficiency, rather than raw intelligence, is now the primary focus, enabling wider adoption of advanced AI models in customer service, research, and content generation. It also pressures competitors to match or surpass these price points, potentially accelerating innovation and price competition in the AI market.

Amazon

AI language model API key

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Background on GPT‑6 and Pricing Trends

OpenAI’s GPT‑6 models, introduced earlier in 2026, marked a step forward in AI capabilities, but their high costs limited widespread use. The company’s recent focus on improving inference efficiency and caching techniques has allowed it to slash prices while maintaining benchmark performance. The release of Sol and Luna follows the Astra variant, which set new standards for model intelligence. Prior to this, GPT‑5.6 models were the industry standard, with higher operational costs that constrained deployment at scale. The new models reflect a strategic shift towards making advanced AI more affordable and accessible, aligning with industry trends toward democratization of AI technology.

Amazon

affordable AI chatbot development tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Remaining Questions About Model Performance and Use Cases

While benchmark scores and hallucination rates are promising, it is still unclear how these models perform in complex, real-world applications beyond testing environments. Some evaluations indicate regressions in knowledge work tasks, suggesting that certain output qualities may be affected by the new tuning. Additionally, long-term stability, user experience, and performance in diverse domains remain to be seen as more organizations adopt these models in production settings. Details about how these models will be integrated into existing workflows and their impact on accuracy in critical tasks are still emerging.

Amazon

GPT-6 compatible AI development software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps in Adoption and Performance Monitoring

OpenAI and early adopters will likely conduct further testing across various industries to validate the models’ reliability and effectiveness in real-world scenarios. Monitoring performance in customer support, content creation, and knowledge work will inform whether the models can replace or supplement existing solutions. Additional updates on model tuning, feature enhancements, and cost management strategies are expected as OpenAI continues refining its offerings. Market competition may also drive other AI providers to release similar cost-effective models, further shaping the AI landscape in the coming months.

Amazon

AI inference optimization hardware

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

How much cheaper are GPT‑6 Sol and Luna compared to previous models?

GPT‑6 Sol and Luna are priced at approximately 50% less than GPT‑5.6 models, with Sol costing $2.00 per 1 million tokens for input and $10.00 for output, and Luna costing $0.10 and $0.50 respectively.

Do the new models perform worse than previous versions?

Benchmark scores indicate performance remains comparable, with Sol scoring 48 and Luna 37 on their respective indices. However, some evaluations show regressions in knowledge work tasks, possibly due to changes in output presentation quality.

What improvements enable the cost reductions?

Enhanced caching and inference techniques, including 90% discounts on cached input reads and better reuse of context, allow these models to be served more efficiently, lowering operational costs.

Will these models be suitable for all AI applications?

While promising for many use cases, some tasks requiring detailed, well-presented outputs might be affected by the models’ tuning for lower cost, so testing in specific workflows is recommended before full deployment.

What is likely to happen next in the AI market?

Further testing, performance validation, and potential new releases from competitors are expected, as the industry moves toward more cost-effective and accessible AI solutions.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

Single Digits: The April That Closed the Open-Weight Gap

In April 2026, the benchmark gap between open-weight and closed proprietary AI models shrank to single digits, reshaping AI economics and strategy.

VigilSAR Benchmark: There Is No Best Model

VigilSAR’s new benchmark reveals that there is no one-size-fits-all model for defense-relevant AI, emphasizing context-specific suitability over raw capability.

OpenAI’s Jalapeño Chip: How Far Has It Really Come In AI?

OpenAI releases initial performance results for its Jalapeño inference chip, highlighting efficiency and architecture innovations, but deployment remains pending.

What Makes Mixture-of-Experts Essential For Frontier AI Growth

Exploring how Mixture-of-Experts models enable scalable, cost-effective AI development by separating memory and compute costs, crucial for future AI advancements.