Why OpenAI’s GPT‑6 Sol And Luna Are Now Half Price But Still Benchmark The Same
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Why OpenAI’s GPT‑6 Sol And Luna Are Now Half Price But Still Benchmark 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 announced that GPT-6 Sol and Luna models are now priced at 50% less than GPT‑5.6 models, while their benchmark scores remain stable. The move aims to expand AI accessibility without sacrificing quality.

OpenAI has officially released its GPT‑6 Sol and GPT‑6 Luna models at half the previous prices, marking a significant shift in AI affordability. The models, introduced on September 22, 2026, are positioned as cost-effective alternatives for a range of applications, with no compromise on benchmark performance.

Both GPT‑6 Sol and Luna now cost approximately 50% less than their GPT‑5.6 predecessors, thanks to improvements in caching and inference efficiency, according to OpenAI. The pricing structure now offers GPT‑6 Sol at $2.00 per 1 million tokens for input and $10.00 for output, down from $4 and $20 respectively, while Luna is priced at $0.10 and $0.50, down from $0.20 and $1.20. Despite the lower prices, independent evaluations by Artificial Analysis show that their benchmark scores remain stable or improved, with GPT‑6 Sol scoring 48 on the Artificial Analysis Intelligence Index, well above the median for comparable models, and Luna scoring 37.

Cost reductions are primarily driven by optimized caching strategies, allowing OpenAI to serve these models at lower operational costs, with cached input reads receiving a 90% discount. The models also demonstrate gains in hallucination reduction, with Sol reducing its hallucination rate from 92% to 60%, and Luna from 93% to 77%. However, some performance regressions were noted in specific knowledge tasks, attributed to changes in presentation quality and output detail, as observed in independent evaluations.

At a glance
updateWhen: announced September 22, 2026
The developmentOpenAI has reduced the prices of GPT‑6 Sol and Luna models by 50% without degrading their benchmark performance, enabling more widespread AI deployment.

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 Management

This price reduction significantly broadens the scope for AI integration in business workflows, research, and customer service, by lowering the financial barrier. Companies can now leverage high-performance models at a fraction of the previous cost, enabling more extensive automation and AI-driven decision-making. The stable benchmark scores suggest that this affordability does not come at the expense of model quality, although some specific knowledge tasks may be affected. Overall, the move underscores a shift toward more accessible AI, potentially accelerating adoption across industries.

Amazon

AI development software tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Background on OpenAI’s Model Pricing and Performance Trends

OpenAI’s recent model releases have focused on balancing performance with cost efficiency. The GPT‑6 family was launched with Astra, a high-end model designed for top-tier results, while Sol and Luna serve as more affordable options. Prior to this, GPT‑5.6 models were priced higher, limiting widespread use in budget-sensitive applications. The release of GPT‑6 Sol and Luna at half the previous prices reflects a strategic emphasis on democratizing AI access, leveraging improvements in inference and caching technologies. Independent evaluations have consistently shown that while higher-end models outperform in complex reasoning, the mid-tier models like Sol and Luna remain competitive for many practical tasks.

Remaining Questions About Model Performance and Use Cases

It is still unclear how these models will perform in long-term, real-world deployments, especially in specialized or high-stakes environments. While benchmarks are stable, some evidence points to regressions in certain knowledge tasks, which could impact workflows requiring detailed, complete outputs. The full impact of reduced presentation quality on production tasks remains to be seen, and user experiences may vary depending on application specifics. Additionally, the long-term durability of the cost savings and caching strategies is yet to be confirmed as usage scales.

Next Steps for Adoption and Evaluation

OpenAI is expected to continue refining these models, with further updates to caching, inference, and tuning based on user feedback. Organizations considering adopting GPT‑6 Sol and Luna should evaluate their specific use cases, especially in tasks requiring detailed output or high accuracy. Monitoring real-world deployments and independent performance assessments will be critical in understanding the models’ full capabilities and limitations. OpenAI may also release further models or updates aimed at optimizing cost and performance balance.

Key Questions

How do GPT‑6 Sol and Luna compare to GPT‑5.6 in terms of performance?

Independent evaluations indicate that benchmark scores for Sol and Luna are comparable to or better than GPT‑5.6, with some regressions noted in specific knowledge tasks. Overall, performance remains stable for most practical applications.

What are the main reasons behind the price reduction?

The price decrease is primarily due to improvements in caching and inference efficiency, which lower operational costs. These savings are passed on to users, making the models more affordable.

Will the lower prices affect the models’ quality or capabilities?

According to OpenAI and independent assessments, the core benchmark performance remains stable or improved. However, some specific quality aspects, like presentation and detailed output, may have been impacted by tuning aimed at reducing hallucinations and improving user experience.

Are there any limitations or risks to using these models at lower cost?

Some knowledge-based tasks may see regressions, and the models’ ability to produce detailed, complete outputs could be affected. Users should evaluate models within their specific workflows to ensure suitability.

What should organizations consider before adopting GPT‑6 Sol and Luna?

Organizations should test the models on their specific tasks, especially those requiring detailed or complex outputs, and monitor ongoing performance as usage scales. Evaluating the impact on quality and cost-efficiency is essential.

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.
EVERGREEN BESTSE

Evergreen bestsellers Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

The Latest In Tech Operations: Apple’s SpeechAnalyzer API Benchmark Analysis

Apple’s new SpeechAnalyzer API has been benchmarked against Whisper and its predecessor, revealing early performance insights for small software teams.

The 10 Biggest AI Advancements To Watch In 2026

Explore the 10 biggest AI advancements expected in 2026, including breakthroughs in natural language processing, autonomous systems, and ethical AI.

Capture The Sky: 9 Best AI Camera Drones For Aerial Video In 2026

Explore the 9 best AI camera drones for aerial video in 2026, highlighting features, capabilities, and what makes each model stand out for filmmakers and hobbyists.

The Strategic Importance Of SeeDance In ByteDance’s AI Expansion Plan

Analysis suggests ByteDance is positioning SeeDance as a frontier AI model, elevating its research arm to compete globally in AI development.