GPT-6 Sol is out, and the price is no longer the same
GPT-6 Sol drops to $2/$10 per million tokens. What it changes for AI costs, automation and vendor choice.
GPT-6 Sol and the new price
GPT-6 Sol is out, and the most visible change is its pricing: $2 per million input tokens and $10 per million output tokens. In practical terms, that cuts the cost in half versus the previous generation and puts it on par with Claude Sonnet 5. For a business, this kind of shift affects more than direct spend: it changes which use cases move from “interesting” to “worth deploying.”
What OpenAI launched
OpenAI introduced GPT-6 Sol and GPT-6 Luna. The business-relevant point is that Sol is the flagship model and its pricing no longer follows the earlier structure. When token costs fall, the question is no longer only whether the AI works, but whether it works at an acceptable margin. That matters most in repetitive workflows, where small savings compound at scale.
Why it matters for AI costs
On a real coding task measured in AutomationBench, cost dropped from about $3.00 to $0.27. That example does not guarantee the same result in every workflow, but it does show a key principle: cost per task can fall materially when the model is cheaper. For finance and operations teams, the right question is whether the savings justify the volume, complexity and human oversight still required.
Business scenarios where the math may change
- Email triage: high-volume work can become more attractive to automate when unit costs fall.
- Document summaries: frequent, standardized tasks are often easier to evaluate on a per-case basis.
- Customer replies: price matters, but so do quality and error risk.
Limitations and risks
Availability is not universal: rollout inside ChatGPT Work and Codex is gradual and only for eligible plans; it is not yet in regular ChatGPT for everyone. Also, a lower price does not remove risks such as incorrect answers, the need for human review or poor integration. And the cheapest model is not automatically the best choice for every task.
Evaluation checklist
- Measure cost per task, not just per token.
- Compare vendors on quality, speed and integration.
- Test with a real workflow before scaling.
- Set human-review thresholds for sensitive cases.
- Recalculate ROI using the new price and your actual volume.
How to apply it in your business
Start with the repetitive processes you already handle today: support, documentation, classification or coding. Then compare the old cost with the new GPT-6 Sol scenario, and decide whether the savings justify automation, expansion or continued oversight. If price is no longer the main barrier, the decision shifts to quality, integration and risk control.