your123 - stock.adobe.com

News

OpenAI takes AI price war to next level with GPT-6 Sol, Luna pricing

The vendor is reacting to U.S. rivals Google and Anthropic, as well as the growing popularity of open source models.

Esther Shittu
By
Published: 23 Sep 2026

OpenAI made an aggressive move in the AI price war by cutting the price of its latest GPT-6 Sol and Luna models by 50% to drive enterprise adoption and compete with rival Anthropic.

The AI lab introduced the models on Sept. 22, revealing that it trained GPT-6 Sol and Luna similarly to how it trained GPT-6 Astra, which it released earlier this month. The AI vendor said that GPT-6 Sol can tackle difficult work tasks, while allowing users to iterate with higher usage limits and lower costs. The model is comparable to Anthropic’s Claude Fable 5.1 for coding at a much lower price, according to OpenAI.

The targeted focus on pricing shows how vendors continue to compete on the price of their new models, as enterprises continually assess how much they’re spending on AI models.

“Companies are so hyper-focused on the economics of running these models, especially the frontier models, because open source and smaller models are [scaring] these larger frontier models hosting providers,” said Bradley Shimmin, an analyst at Futurum Group.

The pressure from open source

With open source vendors from China providing models that provide comparable performance to frontier models, OpenAI, Anthropic and Google have spent the last few months reducing the prices of their models. GPT-6 Sol and Luna highlight just how aggressively OpenAI is willing to compete with Sol. 

The price of GPT-6 Sol is $2 per million input and $10 per million output tokens, which is 50% cheaper than GPT-5.6 Sol. GPT-6 Luna is also half the price of Gpt-5.6 Luna at $0.10 per million input and $0.50 per million output tokens. By comparison, Claude Fable 5.1 costs $10 per million and $50 per million output tokens. Google has also offered promotional pricing on its Gemini 3.8 Flash model till December, with input costing $0.75 and output costing $3.75.

“They’re not just competing with themselves,” Shimmin said. “The downward pressure they’re feeling from the smaller models and open source model movement is real, and I don’t think it’s going to go away. Optimization will continue to be the main differentiator, no longer just the price.”

Pricing and IPO

Reduced pricing not only helps OpenAI compete on cost but also helps it as it pushes toward an IPO, said Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget. While it’s unclear when the model maker will officially go public, it entered its initial filing in June.

“The heavy focus on pricing demonstrates OpenAI's intent to accelerate the adoption of GPT-6 Sol and Luna as fast as possible,” Su said. “It makes a significant difference for enterprises looking to adopt AI in their daily workflow.”

He added that, because the models are smaller versions of GPT-6 Astra, their performance is sufficient to drive AI usage, and, given the price, more enterprises will likely be drawn to them.

Ineffective benefits

However, while price cuts incentivize enterprises, they’re probably unsustainable, and with vendors releasing new models almost every other week, the price benefits appear minimal, Shimmin said. Nevertheless, the price war breeds advancement, he added.

“What we’re getting from this little battle between just a small number of giant closed model makers is an emphasis on what actually matters to the enterprise … that’s innovation,” he said. “It’s not just how good this thing is, it’s how good it is at solving a problem. And if the problem is ‘I can’t afford to run your model,’ they need to solve it. And they’re doing so, and I applaud them for that.”

He added that with GPT-6 Sol and Luna, OpenAI looks more pragmatic, rather than focusing solely on capturing investors' confidence.

“That is the one vote of confidence that actually matters,” Shimmin said. “I don’t care what their valuation is. What I care about is whether they are going to solve a problem for me. Is it going to do so in a secure, performant, governable and trustworthy manner?”

Esther Shittu is an Informa TechTarget news writer and podcast host covering AI and agentic systems.

Dig Deeper on Enterprise AI Strategy