🔍 Read the full analysis: OpenAI Slashes GPT‑6 Sol And Luna Prices By 50% While Benchmark Scores Remain Steady on ThorstenMeyerAI.com
Get movie nights 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 announced a 50% price reduction for its GPT‑6 Sol and Luna models, while their benchmark scores remain stable. This shift aims to make AI more accessible without sacrificing performance, impacting how businesses deploy AI solutions.
OpenAI has announced a 50% reduction in the prices of its GPT‑6 Sol and Luna models, effective immediately. The move comes two weeks after the release of GPT‑6 Astra, and aims to make advanced AI models more affordable for a broader range of users and applications. Despite the price cuts, benchmark scores and performance metrics remain stable, according to independent evaluations, signaling a focus on cost efficiency without sacrificing quality.
On September 22, 2026, OpenAI introduced significant price reductions for its GPT‑6 Sol and Luna models, cutting their costs by 50% compared to previous GPT‑5.6 models. The price of GPT‑6 Sol per 1 million tokens dropped from $4 to $2 for input and from $20 to $10 for output. Similarly, GPT‑6 Luna’s costs fell from $0.20 to $0.10 for input and from $1.20 to $0.50 for output. These reductions are attributed to improvements in caching and inference technologies, which lower operational costs and allow OpenAI to pass savings onto customers.
Independent analysis by Artificial Analysis confirms that these models maintain comparable benchmark scores to their predecessors. GPT‑6 Sol scores 48 on the Artificial Analysis Intelligence Index, well above the median of 25, with a context window of 872,000 tokens and output speed of 115 tokens per second. Luna scores 37, compared to a median of 12, with a 1 million token context window and 154 tokens per second. Cost per task has also been halved, with GPT‑6 Sol at $1.06 and Luna at $0.07, despite slight increases in output tokens per task, indicating that the savings are primarily due to pricing adjustments rather than efficiency gains.
OpenAI emphasizes that these models are designed for cost-sensitive applications, with the Astra model remaining the top choice for tasks requiring maximum quality regardless of cost. The models also show improved hallucination mitigation, with Sol reducing hallucination rates from 92% to 60%, and Luna from 93% to 77%, though at the expense of increased refusals to answer, which may impact certain workflows.
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 of Lower Costs on AI Adoption
The 50% price cut for GPT‑6 Sol and Luna models could significantly expand AI adoption across industries by lowering operational costs. For businesses integrating AI into products or workflows, these models now offer a more affordable way to automate tasks, support research, and enhance customer service. The stable benchmark scores suggest that this cost reduction does not come at the expense of performance, making these models attractive for a broad range of applications. However, the increase in refusals and some regressions in specific knowledge tasks may influence how organizations evaluate these models for critical or detailed work.
As an affiliate, we earn on qualifying purchases.
Background on GPT Model Pricing and Performance
Prior to this announcement, OpenAI’s GPT‑6 models were priced at levels comparable to or higher than previous versions, with GPT‑5.6 models costing $4 per 1 million input tokens and $20 per 1 million output tokens for Sol, and $0.20/$1.20 for Luna. The release of GPT‑6 Astra two weeks earlier marked a new high-performance benchmark, but at a higher cost. The recent price reductions for Sol and Luna reflect a strategic shift to prioritize widespread deployment and cost efficiency. Independent evaluations, such as those from Artificial Analysis, have shown that despite the price cuts, the models’ scores on key AI benchmarks remain stable, indicating that OpenAI’s improvements in caching and inference technology successfully maintain quality while reducing costs.
“The real story here isn’t a smarter model, but a model at half the price, which could radically change what tasks are feasible at scale.”
— Thorsten Meyer, ThorstenMeyerAI.com
cost-effective AI chatbot software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Remaining Questions About Model Performance and Adoption
While benchmark scores remain stable, it is still unclear how these models will perform in long-term or real-world deployments, particularly in complex or sensitive tasks. The impact of increased refusals and the slight regressions in some knowledge evaluations may influence their suitability for certain workflows. Additionally, the full extent of operational savings and how they compare with other AI providers remains to be seen as organizations begin adopting these models at scale.
As an affiliate, we earn on qualifying purchases.
Next Steps for Users and OpenAI’s Strategy
Organizations considering adopting GPT‑6 Sol and Luna should conduct thorough testing to assess performance in their specific use cases, especially for tasks requiring detailed outputs. OpenAI is expected to continue refining these models and may release further updates or new features to optimize cost and quality. Monitoring user feedback and performance metrics over the coming months will be critical to understanding the full impact of these price cuts on AI deployment strategies.
AI performance benchmarking software
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 change over time?
OpenAI has indicated that benchmark scores are stable, but real-world performance may vary depending on specific applications and ongoing model updates.
How will the price reduction affect AI adoption in industry?
The 50% cut could lower barriers for many organizations, enabling broader use of AI for automation, research, and customer support without sacrificing quality.
Are there any trade-offs with the new lower-priced models?
Yes, some regressions in knowledge tasks and increased refusal rates have been observed, which may impact workflows requiring detailed or factual outputs.
What should organizations do before switching to these models?
They should conduct testing specific to their use cases to ensure performance aligns with their needs, especially for critical or detailed tasks.
Will OpenAI release more updates or new models soon?
While not confirmed, OpenAI is likely to continue refining their models and may introduce additional features or versions based on user feedback and technological advances.
Source: ThorstenMeyerAI.com
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
