🔍 Read the full analysis: What OpenAI’s Price Cuts On GPT‑6 Sol And Luna Mean For AI Performance on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
OpenAI has announced a 50% price reduction for GPT‑6 Sol and Luna models, making AI more affordable. While costs have halved, performance metrics show mixed results, with improvements in hallucination reduction but some regressions in knowledge tasks. The move aims to expand AI adoption by lowering operational costs.
OpenAI has announced a 50% price cut for its GPT‑6 Sol and Luna models, effective immediately from September 22, 2026. The models, which are part of the Astra family, now cost half as much as their GPT‑5.6 predecessors, with the company emphasizing that this reduction aims to democratize AI deployment by lowering operational costs for businesses and developers. This move signals a shift from focusing solely on model capabilities to emphasizing cost efficiency as a key driver of AI adoption.
Both GPT‑6 Sol and Luna have been optimized for lower costs through improvements in caching and inference technology, with OpenAI claiming that these enhancements allow the models to be served at significantly reduced expenses. The models’ token prices have been cut by 50%, with GPT‑6 Sol now costing $2.00 per 1 million input tokens and $10.00 per 1 million output tokens, down from $4 and $20 respectively. GPT‑6 Luna’s prices are halved to $0.10 for input and $0.50 for output tokens.
Independent analysis by Artificial Analysis published the same day confirms that the cost per task has roughly halved, with no significant change in overall AI performance scores. GPT‑6 Sol at maximum effort scores 48 on the Artificial Analysis Intelligence Index, well above the median of 25 for models in its class, while Luna scores 37, also above the median. These models demonstrate improved hallucination rates, with Sol reducing hallucinations from 92% to 60%, and Luna from 93% to 77%, primarily by declining to answer more often.
However, some regressions have been observed in knowledge-based tasks, with GPT‑6 Sol dropping around 100 Elo points and Luna about 75 points in evaluations focused on economic and knowledge work. Analysts attribute these regressions to a decrease in presentation quality and deliverable completeness, likely due to tuning adjustments aimed at making models more concise and pleasanter in conversational settings. This suggests a trade-off between improved hallucination management and detailed output quality.
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
Implications for AI Deployment Costs and Quality
The price reductions make AI models more accessible for a broader range of applications, from customer service to research automation. Lower operational costs can enable smaller businesses to adopt advanced AI, potentially accelerating innovation and digital transformation. However, the observed regressions in knowledge accuracy and output completeness highlight that cost-cutting measures may impact the quality of certain AI tasks, especially those requiring detailed, structured information. This development underscores the importance of testing models thoroughly before deploying them in critical workflows, as performance trade-offs may vary depending on use case.
As an affiliate, we earn on qualifying purchases.
Background on GPT‑6 and Cost Optimization Strategies
OpenAI introduced GPT‑6 Astra earlier in 2026, emphasizing high-end capabilities and performance. The new Sol and Luna models, unveiled on September 22, 2026, are positioned as more affordable options aimed at expanding AI adoption. The company attributes the cost reductions to improvements in caching, inference efficiencies, and optimized deployment techniques, which pass savings onto users. This shift reflects a broader industry trend toward balancing model sophistication with operational affordability, especially as AI moves from experimental to operational phases across industries.
Independent evaluations, such as those by Artificial Analysis, have been instrumental in quantifying the impact of these changes on performance metrics, confirming that while costs have decreased significantly, performance remains competitive but with some noted regressions in specific tasks. Prior to this, AI models like Anthropic’s Claude Opus 5.5 also announced price cuts, indicating a competitive landscape focused on affordability.
affordable AI chatbot development tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Aspects of Performance and Use Cases Are Still Unclear
It remains unclear how these models will perform in long-term, high-stakes applications, or whether further tuning will address the observed regressions in knowledge accuracy. The impact on complex, multi-step tasks and real-world deployment stability is still being evaluated. Additionally, the full extent of how caching improvements will influence large-scale deployment costs and efficiency remains to be seen, especially for organizations with diverse workflows.
AI model inference optimization hardware
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for OpenAI and AI Users Post-Price Cut
OpenAI is expected to continue refining its models, potentially releasing further updates to balance performance and cost. Organizations should conduct thorough testing of GPT‑6 Sol and Luna in their specific use cases, especially for tasks requiring high accuracy and detailed outputs. Monitoring the models’ real-world performance and hallucination rates will be crucial as deployment scales. Additionally, competitors may respond with their own price strategies, influencing market dynamics in AI services.
cost-effective AI development 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’s token prices are halved from $4 to $2 for input and from $20 to $10 for output tokens. Luna’s prices are reduced from $0.20 to $0.10 for input and from $1.20 to $0.50 for output tokens, representing a 50% cost reduction across the board.
Does the price cut affect the AI models’ performance?
Independent analysis shows that performance scores remain roughly stable, with some improvements in hallucination reduction but regressions in knowledge-based tasks. Overall, the models are still competitive, but users should evaluate them for their specific needs.
What are the main trade-offs associated with the new models?
While hallucination rates have decreased, some models now produce less detailed or complete outputs, especially in knowledge-intensive tasks. Additionally, there are some regressions in certain economic and knowledge work benchmarks, likely due to tuning for conversational friendliness.
How might these changes influence AI adoption in businesses?
Lower costs make AI more accessible to smaller organizations, potentially accelerating automation and innovation. However, organizations must carefully test models to ensure performance aligns with their operational requirements.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
