🔍 Read the full analysis: What Does The Price Reduction Of GPT‑6 Sol And Luna Mean For AI Users? on ThorstenMeyerAI.com
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
TL;DR
OpenAI has launched GPT‑6 Sol and Luna at half the previous prices, significantly lowering costs for AI applications. While performance remains stable overall, some regressions in knowledge tasks are noted. This shift could expand AI adoption across industries.
OpenAI has introduced GPT‑6 Sol and GPT‑6 Luna at prices approximately 50% lower than their GPT‑5.6 predecessors, effective immediately. The move aims to make AI more accessible for a broad range of applications, from customer support to data analysis, by significantly reducing the cost per task while maintaining comparable performance levels. This development is confirmed by OpenAI’s official statement and independent evaluations published concurrently.
OpenAI’s GPT‑6 Sol and Luna models, released on September 22, 2026, feature a 50% reduction in cost per 1 million tokens compared to GPT‑5.6 versions. For example, GPT‑6 Sol now costs $2.00 per 1M input tokens and $10.00 per 1M output tokens, down from $4 and $20 respectively. GPT‑6 Luna is priced at $0.10 and $0.50, halving previous costs. The company attributes the price drop to improved caching and inference efficiencies, passing savings directly to users.Artificial Analysis’s independent evaluation confirms that costs per task have roughly halved, with GPT‑6 Sol costing about $1.06 per task and Luna approximately $0.07, compared to their GPT‑5.6 counterparts. Despite slightly increased output tokens per task, the savings are primarily due to lower token prices. Performance scores on AI and coding indices remain high, with GPT‑6 Sol scoring 48 and Luna 37 on the Artificial Analysis Intelligence Index, both well above median scores for their class. Hallucination rates have decreased significantly, with Sol’s hallucination rate dropping from 92% to 60%, and Luna’s from 93% to 77%, partly due to models refusing to answer more often.
However, some regressions were observed: GPT‑6 Sol and Luna showed declines in knowledge-based evaluations, such as the GDPval‑AA and AA‑Briefcase benchmarks, attributed to changes in presentation quality and answer completeness. These adjustments may impact workflows requiring detailed, well-structured outputs, especially in document production or comprehensive research tasks.
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.
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.
The effort dial moves cost more than the model choice
| Model and effort | Intelligence Index | Cost 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
How Lower Costs Will Broaden AI Adoption
The price reductions for GPT‑6 Sol and Luna could significantly expand AI deployment across industries by lowering operational costs. For developers and businesses, this means more tasks can be automated at a lower expense, potentially enabling new applications and scaling existing ones more efficiently. While performance remains largely stable, the noted regressions in detailed knowledge tasks suggest some workflows may need testing before full adoption. Overall, the move marks a shift toward making advanced AI accessible to a wider audience, possibly accelerating AI integration in sectors like customer service, research, and automation.
As an affiliate, we earn on qualifying purchases.
Background on GPT‑6 Model Pricing and Performance
OpenAI’s recent model releases have focused on balancing performance improvements with cost efficiency. GPT‑6 Astra, launched two weeks prior, set a new standard for AI intelligence, but its high costs limited widespread adoption. The introduction of GPT‑6 Sol and Luna at half the previous prices aims to democratize access, especially for mid-tier applications. Independent evaluations, such as those by Artificial Analysis, have shown that while costs have dropped, the models maintain strong performance in AI and coding tasks, though some knowledge-related regressions have been observed. The emphasis on caching and inference improvements has been central to these cost savings, reflecting a broader industry trend toward optimizing AI infrastructure for affordability.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Long-term Performance
It is not yet clear how these models will perform over extended use in production environments, especially regarding the noted regressions in knowledge tasks. The impact on workflows that require detailed, structured outputs remains uncertain, and further testing is needed to determine if the regressions are temporary or indicative of broader limitations. Additionally, the effects on user experience in real-world applications, such as customer-facing chatbots or research assistants, are still being evaluated.
As an affiliate, we earn on qualifying purchases.
Next Steps for Adoption and Evaluation
Organizations are advised to conduct pilot tests of GPT‑6 Sol and Luna within their workflows to assess performance, especially for tasks requiring detailed outputs. OpenAI is expected to release further updates and diagnostics tools to help users optimize effort levels and caching strategies. Industry analysts will continue monitoring the models’ long-term reliability and the impact of cost reductions on AI adoption rates, particularly in sectors where budget constraints previously limited AI integration.
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’s input costs are reduced from $4 to $2 per 1M tokens, and output costs from $20 to $10. Luna’s costs are halved from $0.20 to $0.10 for input, and from $1.20 to $0.50 for output, representing roughly 50-60% savings.
Will the performance of GPT‑6 Sol and Luna match GPT‑5.6 models?
Overall, performance remains high, with scores on AI and coding indices comparable to previous models. However, some knowledge-based evaluations show regressions, particularly in detailed presentation quality, which may affect certain workflows.
What are the main benefits of the price reductions?
The primary benefit is increased accessibility for a broader range of applications, enabling more tasks to be automated cost-effectively. This could lead to wider AI adoption in industries like customer service, research, and automation.
Are there any risks or downsides to the new models?
Potential regressions in knowledge accuracy and presentation quality could impact workflows requiring detailed, structured outputs. Users should test the models in their specific use cases before full deployment.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
