OpenAI Cuts GPT-6 Prices by 50% in Escalating AI Model Price War
OpenAI is stepping up the pressure in the enterprise AI market. On September 22, 2026, the company introduced GPT-6 Sol and Luna — two smaller variants of its GPT-6 Astra model — priced at $2 per million input tokens and $10 per million output tokens. That is 50% cheaper than comparable prior offerings.
The move is a direct response to Anthropic’s Claude Fable 5.1 and the broader competitive threat from open-source models out of China that now deliver frontier-level performance at a fraction of the cost.
Key Takeaways
- 50% price cut on GPT-6 Sol and Luna: Input pricing at $2/million tokens, output at $10/million, aimed at accelerating enterprise adoption.
- IPO momentum: OpenAI filed its initial IPO documents in June 2026; lower pricing helps boost the revenue base and user adoption metrics investors will scrutinize.
- Open source is the real pressure: Analysts noted vendors are not just competing with themselves — Chinese open-source models providing comparable capability are forcing all closed-model vendors downward on price.
- Sustainability questions: Experts warn that aggressive price cuts are likely unsustainable, and the real signal to watch is investor confidence, not pricing alone.
GPT-6 Sol and Luna are designed as cost-efficient entry points for companies that need capable but not cutting-edge performance. OpenAI frames the price reduction as a growth lever ahead of its IPO, with the goal of making the economics of running its models work at scale for business customers.
Read the full article on AI Business
Stay in Rhythm
Subscribe for insights that resonate • from strategic leadership to AI-fueled growth. The kind of content that makes your work thrum.
More from Thrum
Additional pieces exploring adjacent ideas
