OpenAI’s Price Reduction For GPT‑6 Sol And Luna: What Remains Unchanged?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: OpenAI’s Price Reduction For GPT‑6 Sol And Luna: What Remains Unchanged? on ThorstenMeyerAI.com

Prime Big Deal Days · Oct 6–7Offer from Amazon

Get the latest gadgets delivered free — and shop member deals

  • Fast, free delivery on millions of items
  • Access to Prime Big Deal Days deals on October 6–7
  • Prime Video, Amazon Music and more included
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

OpenAI has reduced prices for GPT‑6 Sol and Luna models by 50%, aiming to make AI more accessible and cost-effective. The models maintain similar performance levels, with notable improvements in hallucination reduction, though some regressions in knowledge tasks are reported.

OpenAI has introduced a 50% price reduction for its GPT‑6 Sol and Luna models, effective immediately. The move aims to broaden access to advanced AI capabilities by significantly lowering operational costs, making these models more viable for a wider range of applications and organizations.

The new pricing sees GPT‑6 Sol’s input costs drop from $4 to $2 per 1 million tokens, and its output costs from $20 to $10. Similarly, GPT‑6 Luna’s input costs are halved from $0.20 to $0.10, and output costs from $1.20 to $0.50. OpenAI attributes this reduction to improvements in caching and inference technology, which lower operational expenses and allow the savings to be passed on to users.

Independent analysis by Artificial Analysis confirms that the cost per task has roughly halved, with GPT‑6 Sol at maximum effort costing about $1.06 per task—around 50% less than GPT‑5.6 Sol at $1.99—and Luna at $0.07 per task, approximately 60% less than its predecessor. Despite the lower costs, the models’ performance scores remain stable or improve slightly in some areas, especially in hallucination reduction, where Sol’s hallucination rate decreased from 92% to 60%, and Luna’s from 93% to 77%.

However, some areas show regressions. In knowledge-based evaluations, such as GDPval‑AA v2.1 and AA‑Briefcase, both models experienced notable score drops, which the analysis attributes to reduced presentation quality and omission of detailed responses. OpenAI’s own release notes indicate a shift toward shorter, less detailed answers, which might impact workflows requiring comprehensive outputs.

At a glance
updateWhen: announced September 22, 2026
The developmentOpenAI announced on September 22, 2026, that it has halved prices for its GPT‑6 Sol and Luna models, emphasizing increased cost efficiency.

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 price reductions for GPT‑6 Sol and Luna are significant because they lower the financial barrier for deploying large language models in various applications, from customer service to research. This shift can enable smaller organizations to integrate advanced AI, expand automation, and experiment with new use cases without prohibitive costs. While performance remains largely stable, the trade-offs in answer completeness and knowledge accuracy could influence the suitability of these models for certain tasks, especially those requiring detailed, high-fidelity outputs.

Amazon

AI development hardware

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Background on OpenAI’s Model Pricing and Performance

Prior to this announcement, OpenAI’s GPT‑5.6 models were the standard for high-end AI applications, with higher costs limiting broader adoption. The release of GPT‑6 Astra earlier in September introduced a top-tier model with enhanced capabilities, but at a premium price. The new Sol and Luna models are positioned as more affordable options, focusing on cost efficiency rather than pushing the frontier of AI intelligence. The improvements in caching and inference techniques have been key to enabling these lower prices, reflecting ongoing efforts to optimize operational costs in AI deployment.

Independent evaluations have shown that while the new models deliver similar or improved performance in some areas, they also exhibit regressions in others, particularly in producing comprehensive, well-structured outputs. This highlights the ongoing challenge of balancing cost, performance, and output quality in AI model development.

Amazon

GPU inference acceleration cards

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Remaining Questions on Model Performance and Use Cases

It is still unclear how these models will perform in long-term, high-stakes applications, given some observed regressions in knowledge tasks and detailed output quality. The full impact of reduced answer length and omitted details on workflows that depend on comprehensive responses remains to be seen. Additionally, the long-term effects of improved hallucination mitigation versus potential declines in answer completeness are not yet fully understood.

Amazon

AI model deployment tools

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Next Steps for Adoption and Evaluation

Organizations considering adopting GPT‑6 Sol and Luna should conduct thorough testing, especially for tasks requiring detailed, structured outputs. OpenAI is expected to release further updates and diagnostics tools to help users optimize model effort levels and caching strategies. Monitoring real-world performance and user feedback over the coming months will be critical to assess the models’ suitability for diverse applications.

Amazon

cost-effective AI server

As an affiliate, we earn on qualifying purchases.

As an affiliate, we earn on qualifying purchases.

Key Questions

Will the performance of GPT‑6 Sol and Luna match GPT‑5.6?

While initial evaluations show comparable or improved scores in some areas, there are noted regressions in knowledge tasks and detailed output quality. Users should test these models for their specific use cases.

How much cost savings can I expect with these new models?

According to OpenAI, costs are roughly halved: GPT‑6 Sol’s input costs are now $2 per million tokens, and Luna’s are $0.10, with output costs similarly reduced. Independent analysis confirms around 50-60% savings per task.

Are there trade-offs in model quality with the price reduction?

Yes. While hallucination rates are reduced, some evaluations indicate regressions in knowledge and detailed answer quality, especially in producing complete deliverables.

Will these models be suitable for high-stakes applications?

It depends. The models show improved hallucination mitigation but may have limitations in delivering comprehensive outputs. Caution and thorough testing are advised for critical use cases.

What improvements have enabled the price cuts?

Enhanced caching and inference efficiencies have significantly reduced operational costs, allowing OpenAI to pass savings to users without sacrificing model performance.

Source: ThorstenMeyerAI.com

FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

The Future Of Smartwatches In 2026: 10 With Superior AI

Exploring the top 10 smartwatches in 2026 featuring superior AI capabilities, their features, significance, and what to expect next.

Why SAP’s €1 Billion AI Investment Is A Game-Changer For Data Tables

SAP has acquired Prior Labs for over €1 billion to develop advanced tabular foundation models, marking a major shift in enterprise data AI.

Outcome-First Decisions: The Friction Is the Feature

A new decision framework emphasizes testing and evidence, reducing wasted time and money in business choices. Here’s what is confirmed and why it matters.

Launch HN: ProvenMetal (YC S26) Delivers Circuit Boards In Days Instead Of Weeks

ProvenMetal, a YC S26 startup, now delivers assembled circuit boards within days, significantly faster than traditional manufacturing timelines.