🔍 Read the full analysis: OpenAI Slashes GPT‑6 Sol And Luna Prices Without Altering Benchmark Results on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI has slashed the prices of its GPT‑6 Sol and Luna models by half without altering their benchmark performance. This move aims to improve cost-efficiency for AI applications while keeping the models’ capabilities stable. The development could significantly influence AI deployment costs and strategies.
OpenAI has slashed the prices of its GPT‑6 Sol and GPT‑6 Luna models by 50% effective immediately, while maintaining their benchmark performance levels. The move aims to make advanced AI models more accessible for a broader range of applications, emphasizing cost-efficiency without sacrificing quality. This development is significant for companies and developers integrating AI into their workflows, as it could lower operational costs substantially.
On September 22, 2026, OpenAI announced that both GPT‑6 Sol and Luna models now cost half their previous prices, with GPT‑6 Sol reducing from $4 to $2 per 1 million input tokens and from $20 to $10 per 1 million output tokens. Similarly, GPT‑6 Luna’s prices dropped from $0.20 to $0.10 for input and from $1.20 to $0.50 for output tokens. These reductions are attributed to improvements in caching and inference techniques, which allow OpenAI to serve these models at lower costs, passing savings onto users.
Independent analysis by Artificial Analysis confirmed that the models’ benchmark scores remained stable or improved slightly, despite the significant price cuts. GPT‑6 Sol scored 48 on the Artificial Analysis Intelligence Index, well above the median of 25, with a context window of 872,000 tokens. Luna scored 37, also above the median of 12, with a 1 million-token context window. Cost per task dropped by approximately 50-60%, even as the models produced more output tokens per task, indicating the savings stem from price reductions rather than efficiency gains.
OpenAI emphasizes that these models are designed for cost-sensitive applications, with Astra remaining the top-tier option for highest-quality results regardless of expense. The models also demonstrate improvements in hallucination rates, with Sol reducing hallucinations from 92% to 60%, and Luna from 93% to 77%, partly by declining to answer more often, which lowers errors but also reduces the number of responses.
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
Impact on AI Deployment Costs and Strategies
The price reductions for GPT‑6 Sol and Luna could dramatically lower the cost barriers for deploying advanced AI models in various industries, from customer service to research. Companies can now access high-performance models at a fraction of previous costs, enabling broader adoption and experimentation. However, the models’ slight regressions in some knowledge tasks suggest users should evaluate their specific workflow needs carefully, particularly if high-quality, detailed outputs are required.
This move underscores OpenAI’s focus on improving cost-efficiency and scalability, potentially reshaping competitive dynamics in the AI market. As more organizations adopt these cheaper models, the overall landscape of AI-powered automation and decision-making may shift towards more affordable, widespread use.
AI language model API pricing plans
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background on GPT‑6 Model Pricing and Capabilities
OpenAI introduced GPT‑6 Astra earlier in September 2026, emphasizing its superior intelligence and capabilities. The company positioned Astra as the flagship model, with Sol and Luna serving as more affordable options designed to democratize access to advanced AI. Prior to these price cuts, GPT‑6 models were priced similarly to GPT‑5.6, making cost a significant barrier for some users.
The recent announcement reflects a strategic shift, leveraging improvements in caching and inference to lower operational costs. This aligns with OpenAI’s broader goal of expanding AI accessibility while maintaining high benchmark performance, as demonstrated by independent evaluations that show stable or improved scores despite the price reductions.
Previously, OpenAI’s pricing model limited some applications due to high costs, especially for large-scale deployments. The new prices could enable more extensive experimentation, integration, and use-case diversification, especially for smaller organizations or those with tight budgets.
Remaining Questions About Model Capabilities and Usage
While benchmark scores remain stable or improved, it is still unclear how these models perform across all real-world tasks, especially those requiring detailed or nuanced outputs. The observed regressions in some knowledge benchmarks suggest potential trade-offs in presentation quality, which could affect use cases demanding high-quality deliverables. Additionally, the long-term impact of reduced hallucination rates and refusal behaviors on user experience warrants further observation.
OpenAI has not disclosed detailed internal metrics or how these price reductions might influence future model updates or the development of Astra’s top-tier capabilities. It is also uncertain whether similar cost reductions will extend to other models or future releases.
Next Steps and Monitoring Developments
OpenAI is expected to continue refining its caching and inference techniques to further reduce costs and improve efficiency. Users should monitor performance evaluations and real-world testing to assess whether the models meet their specific needs, especially in tasks requiring detailed, high-quality outputs.
Further independent assessments and user feedback will likely emerge over the coming weeks, clarifying how these models perform in diverse applications. OpenAI may also announce additional updates or new models designed to optimize the balance between cost and capability in the near future.
Key Questions
How much have GPT‑6 Sol and Luna prices been reduced?
GPT‑6 Sol’s prices are now approximately half of their previous levels, with input costs dropping from $4 to $2 per 1 million tokens and output from $20 to $10. Luna’s prices fell from $0.20 to $0.10 for input and from $1.20 to $0.50 for output tokens, representing about a 50-60% reduction overall.
Do the price reductions affect the models’ performance?
No, independent analysis confirms that benchmark scores and capabilities remain stable or slightly improved, despite the lower prices. However, some knowledge benchmarks showed regressions, possibly due to changes in presentation quality.
What technical improvements enabled the price cuts?
OpenAI cites enhancements in caching and inference techniques, allowing more efficient reuse of context and reducing operational costs, which are then passed on as lower prices for users.
Are there any trade-offs with these new models?
Yes, some regressions in knowledge tasks and a tendency for the models to decline answering more often—reducing hallucinations but also decreasing the number of responses—may impact certain applications requiring detailed outputs.
Will other models also see price cuts?
OpenAI has not announced plans for further price reductions across all models, but ongoing improvements in efficiency suggest future adjustments could occur as technology advances.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
