OpenAI and Anthropic Just Slashed Prices on the Same Afternoon. Here's Why That's a Big Deal

On September 22, Anthropic released Claude Opus 5.5. About 90 minutes later, OpenAI released GPT-6 Sol and Luna. Both companies cut prices hard on the same day, and neither seemed to be waiting on the other. That timing was not an accident. It was a shot fired in a price war that's been building for months.

Here's what actually happened.

Anthropic went first

Anthropic dropped Claude Opus 5.5 as the first model in its new 5.5 family. The headline claim: it costs 40% less to run than Opus 5 on typical workloads, while performing close to their top-tier Fable 5.1 model on most tasks. List price dropped from $5 and $25 per million tokens down to $4 and $20. Output got roughly 30% faster. Anthropic also scrapped the five-hour usage caps for Pro, Max, Team, and Enterprise subscribers, which had been a real annoyance for anyone running AI all day.

OpenAI answered within the hour

OpenAI followed with GPT-6 Sol and Luna, cutting API prices in half compared to the previous 5.6 versions. Sol dropped to $2 and $10 per million tokens, down from $4 and $20. Luna, the smaller and faster model, dropped to $0.10 and $0.50. OpenAI confirmed these are permanent price cuts, not a limited-time promo. The company also claims Sol makes about half as many factual errors as its predecessor.

Both companies say the savings come from real efficiency gains in caching and inference, not just a discount to win headlines.

Why this matters more than another spec sheet

It's easy to read "prices went down" and shrug. But zoom out and the pattern is the real story. Two of the most important AI companies in the world just cut costs by 40 to 50% within the same afternoon, and neither treated it as a temporary sale. That tells you where this industry is heading:

• The cost of running AI at a professional level is dropping fast, not slowly.
• Competition between OpenAI and Anthropic is now openly about price, not just capability.
• What was expensive six months ago is now priced for everyday business use.

This is the same pattern we've seen with every major technology curve. The capability shows up first, expensive and limited. Then the price falls, and it becomes something every business can actually use.

What it means for a growing business

If you've been holding off on AI because it felt like a big expensive bet, that math just changed again. Running an AI agent that handles customer calls, follows up on leads, or processes paperwork all day used to carry real token costs that added up. Those costs just dropped by roughly half, on both of the two leading platforms, in the same week.

That means:

• Systems that were borderline on cost now clear the bar easily.
• Agents that run constantly, not just occasionally, are more affordable to operate.
• The businesses that build now get more compute for their money than the businesses that wait six more months to "see how it plays out."

The caveat worth knowing: lower per-token pricing doesn't automatically mean a cheaper bill. Some tasks use more tokens to get the same result, so the real savings depend on how a system is actually built and run. This is exactly why the setup matters as much as the sticker price.

The takeaway

AI is getting cheaper and more reliable at the same time, and that trend is not slowing down. Two rivals just proved it in the same afternoon. For a growing business, that's not a headline to skim past. It's a signal that the cost of doing nothing keeps going up, while the cost of building something real keeps going down.