AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: OpenAI’s Cost-Cutting For GPT‑6 Sol And Luna Leaves Benchmark Scores Unchanged 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 launched GPT-6 Sol and Luna models at half the price of their GPT-5.6 predecessors, with benchmark scores unchanged. Cost reductions are achieved through improved caching and inference, but some quality regressions in knowledge tasks are noted.

OpenAI has introduced GPT-6 Sol and Luna on September 22, 2026, with prices halved compared to previous models, while benchmark scores remain largely unchanged. This move emphasizes cost efficiency over new capabilities, potentially broadening AI adoption across industries.

Both models are now priced at approximately 50% of their GPT‑5.6 predecessors, 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, according to OpenAI. The price reductions result from improvements in caching and inference technology, which allow OpenAI to serve these models more cheaply while passing savings to users.

Independent evaluation by Artificial Analysis confirms that, despite the lower costs, benchmark scores have remained stable. GPT‑6 Sol scores 48 on the Artificial Analysis Intelligence Index, significantly above the median of 25 for comparable models, with GPT‑6 Luna scoring 37, above a median of 12. These scores reflect maintained or improved capabilities in general intelligence tasks.

However, some regressions are noted in specific knowledge-based evaluations. GPT‑6 Sol experienced a decline of about 100 Elo points in economic task assessments, and Luna about 75 points, with similar drops in multi-week knowledge work benchmarks. The cause appears linked to a shift in model tuning aimed at reducing presentation quality and output length, which may affect detailed deliverables.

At a glance
updateWhen: announced September 22, 2026
The developmentOpenAI announced the release of GPT-6 Sol and Luna models on September 22, 2026, with significant price reductions but stable benchmark performance, affecting AI cost-efficiency and deployment strategies.

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 Cost Reduction on AI Deployment

The price cuts make AI models more accessible for a broader range of applications, especially where cost constraints previously limited deployment. While benchmark scores remain stable, the noted regressions in some knowledge tasks suggest that users should evaluate whether the models meet their specific quality requirements, especially for detailed or complex outputs.

This development could accelerate AI adoption in industries like customer service, research, and automation, where cost efficiency is critical. The trade-offs between cost and quality will influence how organizations integrate these models into their workflows and products.

Amazon

AI model caching hardware

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Background on GPT‑6 and Model Cost Strategies

OpenAI’s GPT‑6 models, Astra, Sol, and Luna, were released in September 2026, following the Astra flagship, which emphasized high performance. The new models focus on delivering similar capabilities at significantly lower prices, achieved through technical improvements such as enhanced caching and inference efficiencies. Prior to this, OpenAI’s pricing for GPT‑5.6 models ranged from $4 to $20 per million tokens, depending on the use case.

The launch reflects a strategic shift toward prioritizing cost-effective AI solutions, aiming to democratize access and enable broader deployment across industries. Independent evaluations have been closely tracking performance and cost metrics to assess the real-world impact of these changes.

Amazon

AI inference optimization tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Remaining Questions on Model Capabilities

It is still unclear how these models will perform in real-world, complex knowledge tasks over longer periods or in specialized domains. The regressions in some benchmarks suggest that further tuning or updates may be needed to optimize for specific use cases. Additionally, the long-term impact of reduced output quality on user satisfaction and trust remains to be seen.

Amazon

cost-effective AI deployment solutions

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for OpenAI and Users

OpenAI is expected to continue refining GPT‑6 models, potentially addressing the identified regressions through updates or new tuning strategies. Users should evaluate these models in their specific workflows, especially for tasks requiring detailed or high-precision outputs. Monitoring real-world performance and benchmarking will be essential to determine the models’ suitability for different applications.

Amazon

AI model benchmarking software

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 costs approximately $2.00 per 1 million input tokens and $10.00 per 1 million output tokens, roughly half the price of GPT‑5.6 Sol. Luna costs $0.10 and $0.50 respectively, about 50–60% less than its predecessor.

Do the new models perform better or worse than previous versions?

In general benchmark tests, GPT‑6 Sol and Luna scores remain stable or improved, but some knowledge-based evaluations show regressions, likely due to tuning for shorter, more concise outputs. Overall, scores indicate maintained or enhanced capabilities in broad intelligence tasks.

What are the main technological improvements enabling cost reductions?

OpenAI attributes the cost savings primarily to advancements in caching and inference technology, which allow more efficient reuse of context data, reducing the computational costs associated with serving the models.

Will the quality of outputs in specific tasks decline?

Some evaluations suggest a decline in detailed knowledge tasks, especially for complex or lengthy outputs, due to tuning for shorter, more user-friendly responses. Users should test the models for their particular needs before full adoption.

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

Upgrade Your Storage With AI: Top 10 NAS Devices In 2026

Discover the best NAS devices of 2026, featuring AI-enhanced storage, performance, and expandability for home and business users.

Top 10 AI Advances That Will Define 2026

A comprehensive overview of the key AI innovations expected to shape 2026, including breakthroughs in natural language processing, autonomous systems, and ethical AI.

9 Best 4K Monitors for Work and Play in 2026

Discover the best 4K monitors of 2026 for productivity and gaming, including features, prices, and what fits your needs best.

The Key Principles Behind Successful AI Context Stack Auditing

An analysis of key principles behind successful AI context stack auditing, based on recent insights from Anthropic’s model optimization practices.