OpenAI Priced Its Model at Half of Anthropic's

The Same Day, Two Price Cuts

On September 22, 2026, Anthropic and OpenAI both cut the price of their newest frontier models within hours of each other. Anthropic's Claude Opus 5.5 dropped to $4 per million input tokens and $20 per million output tokens, down 20 percent from Opus 5's $5 and $25, with cache reads down 60 percent to $0.20. Anthropic says the model runs on typical workloads for 40 percent less than Opus 5 and generates output more than 30 percent faster.

OpenAI answered with GPT-6 Sol and GPT-6 Luna. Sol fell to $2 per million input tokens and $10 per million output tokens, half of the GPT-5.6 Sol price. Luna, the smaller everyday model, dropped to $0.10 and $0.50, a cut of 50 to 58 percent. An OpenAI spokesperson confirmed to VentureBeat that both rates are permanent list prices, not a launch promotion.

Sol Is Priced at Anthropic, Not at Itself

The coincidence stops at the calendar. GPT-6 Sol's new $2 and $10 rate is not just lower than Opus 5.5, it is exactly half of it, and it matches what Anthropic already charges for its own cheaper Sonnet 5 tier. OpenAI did not price Sol against its own predecessor and stop there. It priced Sol against a specific rival product, on the day that product's own price changed.

That is a different kind of announcement than a routine generational price cut. It reads as OpenAI telling enterprise buyers, in dollar terms, exactly where it wants Sol to sit relative to Anthropic's lineup: below the flagship, level with the mid-tier, and cheaper than both on a per-task basis if OpenAI's own benchmark holds up.

The Benchmark Comparisons Are Already Out of Date

OpenAI's highlighted benchmark claim is that GPT-6 Sol scores 33.2 percent on AutomationBench 1.0.6 for $0.27 per task, versus Claude Opus 5 at 11.1 times the cost per task. That comparison uses Opus 5, the model Anthropic just discontinued in favor of Opus 5.5 at 40 percent lower cost. The real per-task gap after September 22 is narrower than the headline figure, though OpenAI has not published an updated comparison against Opus 5.5 specifically.

Anthropic's own marketing has the same problem in reverse: it says Opus 5.5 beats GPT-6 Astra, OpenAI's separate top-tier model launched earlier in September, at 20 to 40 percent of the cost. Neither company's published comparison yet accounts for both companies' same-day cuts, because both cuts landed at the same moment the comparisons were written.

What a Buyer Should Actually Do

Any team with a fixed monthly budget for agentic coding or high-volume model calls just watched its two main options reprice inside the same 24 hours, against each other rather than against their own last generation. The useful move is not to read either press release's benchmark table at face value. It is to re-run your own top workload against GPT-6 Sol and Claude Opus 5.5 at the new prices before renewing any enterprise contract signed even a few weeks ago, because both companies' own comparisons were already stale on the day they were published.

The timing carries one more detail worth keeping in view. This pricing exchange landed ten days after a federal antitrust complaint accused Anthropic, OpenAI, Google and SpaceXAI of coordinating to slow AI development and reduce subscriber value. Two of the four defendants spent September 22 undercutting each other's prices by as much as 58 percent, which is the opposite of the restraint the lawsuit alleges.

Read next: A Safety Essay Became Evidence in an Antitrust Suit  |  Claude Built the Exploit OpenAI Never Patched