OpenAI's Price War Just Cut Your AI Bill

23 September 2026

OpenAI has released two new models, GPT-6 Sol and GPT-6 Luna, alongside improvements to prompt caching that cut latency and cost for developers. On the same day, Anthropic launched Claude Opus 5.5. Both moves land in the middle of what the Financial Times calls an intensifying price war between the two firms. OpenAI also published a case study showing that research firm Parallel cut both time and cost in half after moving to the newer GPT-6 Astra model.

None of this matters to you directly as a business owner, until you realise that most of the AI-powered tools you already pay for, quoting software, chatbots, RAMS drafters, invoice checkers, are built on top of these exact models. When OpenAI or Anthropic cuts the price per query, the cost of running your tool drops too. The question is whether your supplier passes that saving on, or quietly pockets it.

A worked example

Say you pay £150 a month for an AI-powered RAMS drafting tool billed through an API. If the underlying model cost has genuinely halved, as Parallel's case study shows for research tasks, your supplier's margin on that £150 has roughly doubled overnight. If competition in the market holds, prices should follow the cost curve down over the next few renewal cycles. Over a year, a proportional saving on that single tool could be worth several hundred pounds, enough to cover a day's labour or a fresh set of PPE.

The same logic applies at the API level if you or your developer call these models directly. GPT-6 Sol and Luna are priced differently to balance capability against cost, so a task that doesn't need the most powerful model can now run on a cheaper one without losing accuracy. That is a direct lever on your software bill, not a theoretical one.

What to do this month

Prompted by: https://openai.com/index/introducing-gpt-6-sol-and-luna

Want this level of clarity on your own numbers?

Start with the free Margin vs Volume calculator — 30 seconds, no sign-up.

Run your numbers →