← All postsSeptember 28, 2026

OpenAI and Anthropic Both Cut AI Prices This Week — What It Means for Your Budget

On September 22, OpenAI and Anthropic both shipped new flagship models on the same day, and both releases came with genuine price cuts, not just better benchmarks. OpenAI's new GPT-6 Sol and Luna models landed at roughly half what the prior generation cost per million tokens, and the company told VentureBeat the lower rate is permanent, not a limited-time promotion. Anthropic's Claude Opus 5.5 came in around 20% cheaper on list price, and the company says it works out closer to 40% cheaper on typical workloads once caching and speed improvements are factored in.

Neither release is the interesting part by itself — model version numbers roll over every few weeks now, and most small businesses will never notice the difference between one frontier model and the next. What's worth noticing is that the price dropped, twice, from two competing labs, in the same week.

Why the price trend matters more than the model

If you're paying a vendor for an "AI-powered" feature — a chatbot, a document summarizer, an automated triage tool — some real share of that bill covers the AI usage running underneath it. When the underlying cost of running these models keeps falling this fast, one of two things should be happening to what you pay: it should be dropping too, or the vendor is quietly keeping the difference. That's worth asking about, especially for a tool you set up a year ago and haven't looked at since.

What I'd actually do about it

  • If you're on a flat-rate AI tool subscription that hasn't gotten cheaper or better in the last six months, ask the vendor why — or start pricing alternatives.
  • If you're paying for AI usage directly through an API or a custom integration, check whether it's still pointed at an older model. Costs don't fall on their own; someone has to update the configuration to use the newer, cheaper version.
  • Don't chase every price drop by switching vendors constantly — the churn usually costs more than the savings. But a move this size, from two labs at once, is a reasonable prompt to check your own numbers rather than assume AI tooling only gets more expensive as it gets more capable.

The trade press ran with benchmarks and context-window sizes. For a small business, the more useful takeaway is that the tools most of us already build on or buy into got structurally cheaper this week — and that's been a fairly consistent pattern through the year, not a one-off.

Sources: