OpenAI Slashes GPT‑6 Sol And Luna Prices—Benchmark Scores Remain Constant
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: OpenAI Slashes GPT‑6 Sol And Luna Prices—Benchmark Scores Remain Constant on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

  • Fast, free delivery for dorm and study essentials
  • Prime Video and Amazon Music included
  • Member-only deals
Try Prime for Young Adults Free trial for eligible 18–24 year olds
As an affiliate, we earn on qualifying purchases.

TL;DR

OpenAI has announced a 50% price reduction for GPT‑6 Sol and Luna models, effective September 22, 2026, without significant changes to their benchmark scores. This shift aims to make AI more affordable for broader use, though some performance regressions are noted.

OpenAI has reduced the prices of its GPT‑6 Sol and Luna models by approximately 50%, effective September 22, 2026, with no corresponding changes to their benchmark scores. This move aims to make advanced AI more accessible and affordable for a wide range of applications, from enterprise workflows to consumer products.

Both models, GPT‑6 Sol and GPT‑6 Luna, now cost half of what their GPT‑5.6 predecessors did, with GPT‑6 Sol priced at $2.00 per 1 million tokens for input and $10.00 for output, and GPT‑6 Luna at $0.10 and $0.50 respectively. These reductions are attributed to improvements in caching and inference technology, which allow OpenAI to serve these models at lower costs, passing savings directly to users.

Despite the price cuts, independent evaluations by Artificial Analysis show that the models’ benchmark scores have remained roughly constant. GPT‑6 Sol scores 48 on the Artificial Analysis Intelligence Index, well above the median of 25 for similar models, while GPT‑6 Luna scores 37, also above the median of 12. These scores suggest that the models’ core capabilities have not diminished despite the lower prices.

However, some performance regressions have been observed. Notably, GPT‑6 Sol experienced a drop in knowledge evaluation scores, with a decrease of about 100 Elo points on the GDPval‑AA v2.1 benchmark. Similarly, Luna’s scores declined slightly in knowledge work assessments, which the analysis attributes to reduced presentation quality and omitted details, possibly linked to tuning for more concise responses.

At a glance
breakingWhen: announced September 22, 2026
The developmentOpenAI has slashed the prices of GPT‑6 Sol and Luna models by half, maintaining their benchmark scores, which could significantly influence AI deployment costs and applications.

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.

Impact of Price Reduction on AI Deployment Costs

The price cuts for GPT‑6 Sol and Luna significantly lower the cost barrier for integrating advanced AI into products and workflows. This shift enables smaller organizations and startups to access high-performance models previously limited by expense, potentially accelerating AI adoption across industries. While benchmark scores indicate stable core capabilities, the observed regressions in certain knowledge tasks highlight the importance of testing models within specific use cases before large-scale deployment.

Overall, the move emphasizes cost efficiency as a primary driver in AI development, with OpenAI positioning itself to maintain competitive advantage by making high-quality models more affordable without sacrificing performance.

Amazon

AI model cost management tools

As an affiliate, we earn on qualifying purchases.

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 costing more due to enhanced performance. The release of GPT‑6 Astra earlier this month marked a step forward in model intelligence, but the focus on Astra’s capabilities overshadowed the significance of cost reductions in the broader model family.

Two weeks prior, OpenAI introduced GPT‑6 Astra, emphasizing improved performance and new features. The subsequent release of GPT‑6 Sol and Luna on September 22, 2026, shifts the narrative toward cost efficiency, with prices cut by 50%. These models are now positioned as more accessible options for automation and integration, especially for tasks where high accuracy and large context windows are essential.

Independent evaluations, such as those by Artificial Analysis, have provided early insights into the models’ performance, confirming that while costs have decreased, core benchmark scores remain stable. This aligns with OpenAI’s claim that technological improvements in caching and inference enable lower prices without sacrificing quality.

Amazon

AI deployment cost optimization software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Uncertainties About Long-Term Performance and Adoption

It is still unclear how these models will perform across a broader range of real-world tasks over time, especially in complex knowledge work or nuanced customer interactions. The observed regressions in some benchmarks raise questions about potential trade-offs made during tuning for cost efficiency. Additionally, the long-term impact on market competition and adoption rates remains to be seen as users evaluate the models in diverse settings.

Amazon

AI token counting tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for Evaluation and Market Adoption

OpenAI is expected to continue monitoring model performance through ongoing evaluations and user feedback. Further independent testing will clarify how these models perform in different operational contexts. Meanwhile, organizations interested in adopting these models should conduct their own testing to verify suitability for their specific use cases, especially in applications requiring high accuracy and detailed outputs.

OpenAI may also release updates or tuning adjustments based on early feedback, aiming to balance cost savings with performance stability. The broader AI community will likely scrutinize these developments, assessing whether the cost reductions lead to wider adoption and innovation.

Amazon

AI model benchmarking tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

What are the main changes in GPT‑6 Sol and Luna?

OpenAI has halved the prices of GPT‑6 Sol and Luna models, primarily through improvements in caching and inference technology, while benchmark scores remain roughly the same as previous models.

Do the lower prices affect the models’ performance?

According to independent evaluations, core benchmark scores are stable, but some knowledge and quality regressions have been observed, particularly in detailed or complex tasks.

Why did OpenAI reduce the prices now?

The reduction is driven by technological advances that lower operational costs, enabling OpenAI to pass savings to users and broaden AI accessibility.

Will these models be suitable for all applications?

While cost-effective, organizations should test these models within their specific workflows, especially for tasks requiring high accuracy and detailed outputs, due to some noted regressions in certain benchmarks.

What is likely to happen next?

OpenAI will continue evaluating performance and gathering user feedback, potentially releasing further updates. Market adoption will depend on how well these models perform in real-world applications.

Source: ThorstenMeyerAI.com

EVERGREEN BESTSE

Evergreen bestsellers Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

Virginia State University Surges In Global Coverage

Virginia State University experiences a significant surge in global media mentions, with 13 reports in a recent window, highlighting increasing international recognition.

The license. Why the AI content market pays the brand-name corpus and strands the long tail.

Analysis of how licensing favors large publishers, marginalizing small sites and the potential for collective licensing to address this imbalance.

The Door: Why the Interface Is Worth More Than the Model

SpaceX acquired a $60 billion coding interface, highlighting the growing importance of the user interface over AI models in control and distribution.

Formalizing Fermat’s Last Theorem

Researchers have announced a formal proof of Fermat’s Last Theorem, marking a major milestone in mathematical rigor and proof verification.