GPT‑6 Sol And Luna Prices Cut In Half, AI Benchmarks Remain Consistent
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: GPT‑6 Sol And Luna Prices Cut In Half, AI Benchmarks Remain Consistent 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 cut the prices of GPT‑6 Sol and Luna models by half, while their AI benchmark scores remain stable. The move aims to make advanced AI more accessible without sacrificing performance, but some quality regressions are noted.

OpenAI has significantly reduced the prices of its GPT‑6 Sol and Luna models by 50%, with Sol now costing $2.00 per 1 million input tokens and Luna $0.10, compared to previous prices of $4 and $0.20 respectively. Despite the price cuts, benchmark evaluations indicate that the models’ AI performance remains consistent, with no notable decline in core capabilities. This move aims to broaden access to advanced AI by making cost-effective options more widely available, impacting how businesses and developers deploy large language models in various workflows.

On September 22, 2026, OpenAI introduced a price reduction for its GPT‑6 Sol and Luna models, now available at half their previous costs. GPT‑6 Sol’s input costs dropped from $4 to $2 per 1 million tokens, and Luna’s from $0.20 to $0.10, with output prices also halved. The reduction is driven by improvements in caching and inference efficiencies, allowing OpenAI to serve these models more cheaply while passing savings to users.

Independent analysis by Artificial Analysis confirms that the models’ benchmark scores remain stable, with Sol achieving 48 points and Luna 37 on the Artificial Analysis Intelligence Index, well above their respective medians in their price classes. Cost per task has also halved, with Sol at maximum effort costing approximately $1.06 per task, and Luna at about $0.07, representing roughly 50-60% savings over previous models. Notably, these savings come despite a slight increase in output tokens per task, indicating cost reductions are primarily from pricing and efficiency gains rather than improved performance.

However, some quality regressions are observed. Artificial Analysis reports declines in certain knowledge-work evaluations, such as GDPval‑AA v2.1 and AA‑Briefcase, where models scored lower than previous versions. These regressions are attributed to reduced presentation quality and omitted details, suggesting that while models are cheaper and generally stable, their output may be less comprehensive in some contexts. OpenAI’s internal notes acknowledge fewer low-value details and shorter answers, which could impact workflows requiring detailed deliverables.

At a glance
updateWhen: announced September 22, 2026; ongoing i…
The developmentOpenAI announced a 50% price reduction for GPT‑6 Sol and Luna models on September 22, 2026, with benchmarks showing stable performance despite lower costs.

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 Lower Costs on AI Deployment Strategies

The 50% price reduction for GPT‑6 Sol and Luna models significantly lowers the financial barrier for deploying advanced AI across industries. This shift can enable smaller companies and startups to incorporate powerful language models into their products and services, expanding AI’s reach into new markets. Additionally, the stable benchmark performance suggests that users can adopt these models without fearing a loss in quality, making AI-driven automation more feasible and cost-effective. However, the noted regressions in detailed output quality highlight the importance of testing these models within specific workflows to ensure they meet quality standards.

Amazon

AI language model API subscription

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Background on GPT‑6 Model Pricing and Performance

OpenAI’s GPT‑6 models, Astra, Sol, and Luna, were launched in September 2026 as part of the Astra family, emphasizing cost efficiency alongside high performance. Previously, GPT‑6 Sol and Luna models were priced at $4 and $0.20 per 1 million tokens for input, and $20 and $1.20 for output, respectively. The models are designed with improvements in caching and inference, allowing for significant cost reductions. Independent evaluations by Artificial Analysis, published concurrently, confirmed that despite the price cuts, the models’ benchmark scores remained stable, with Sol outperforming median scores for its class.

The move follows a broader industry trend of reducing AI model costs, exemplified by competitors like Anthropic, which cut prices of their models. OpenAI’s focus on balancing cost and performance aims to democratize access to advanced AI while maintaining quality, though some trade-offs in output detail have been observed in recent evaluations.

Amazon

large language model development tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Remaining Questions About Model Quality and Use Cases

It is not yet clear how widespread the quality regressions are across different workflows, especially for tasks requiring detailed, nuanced outputs. OpenAI acknowledges shorter and less detailed answers but has not specified whether these regressions impact all use cases equally. The long-term stability of benchmark scores and real-world performance in diverse applications remain to be seen, as ongoing evaluations are underway.

Amazon

AI model deployment software

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Future Steps for OpenAI and AI Market Adoption

OpenAI is expected to continue refining its caching and inference techniques to further reduce costs and improve output quality. Industry analysts anticipate more model updates and price adjustments in the coming months as competing firms also pursue cost reductions. Users should monitor OpenAI’s ongoing performance reports and conduct workflow-specific testing before fully deploying these models at scale. Additionally, further independent evaluations are likely to clarify the models’ suitability for different tasks, especially those demanding high-detail outputs.

Amazon

cost-effective AI development platform

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

Why did OpenAI reduce the prices of GPT‑6 Sol and Luna models?

OpenAI reduced prices by approximately 50% due to improvements in caching and inference efficiency, which lowered operational costs while maintaining model performance.

Do the benchmark scores indicate that the models are less capable now?

No, independent evaluations show that benchmark scores remain stable, indicating that core AI capabilities are maintained despite the lower cost structure.

Are there any downsides to the price cuts?

Some evaluations reveal regressions in output detail and presentation quality, which could affect workflows that depend on comprehensive, well-structured responses.

How might this impact AI adoption in industry?

The significant cost reductions could enable broader adoption of advanced AI, especially among smaller organizations, increasing automation and AI-driven innovation across sectors.

What should users do before switching to these models?

Users should test the models within their specific workflows to ensure output quality meets their standards, particularly for tasks requiring detailed and nuanced responses.

Source: ThorstenMeyerAI.com

FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

When One Agent Isn’t Enough: Claude Now Builds Its Own Team Of Agents On The Fly

Anthropic’s Claude now autonomously builds and orchestrates teams of agents for complex tasks, enhancing multi-step workflows and accuracy.

Maximize AI Performance With These 9 Mobile Workstations In 2026

Discover the nine best mobile workstations in 2026 optimized for AI performance, balancing power, portability, and reliability for professionals.

IdeaClyst: The Validation Council

IdeaClyst, an idea validation tool using model cross-examination, has officially launched, promising more rigorous idea stress-testing for businesses.

The Future Of Work: 11 Best AI Productivity Software In 2026

Discover the 11 best AI productivity tools in 2026, including Microsoft Copilot and Claude AI, and understand how they transform work workflows today.