OpenAI and Anthropic in price war as Chinese AI rivals gain ground

Price War Erupts as Chinese AI Rivals Reshape the Market

Chinese AI rivals are forcing OpenAI and Anthropic into a brutal price war. Both US labs are slashing costs on mid-tier models, a desperate bid to hold onto budget-conscious customers who've grown weary of ballooning AI bills. So the defections are real. Corporate clients, hit with rising costs, increasingly jump ship to cheaper alternatives from Chinese developers like Moonshot and DeepSeek, and that bleeding won't stop anytime soon because the price gap keeps widening. But it's a fight they can't easily win.

OpenAI just slashed GPT-5.6 Luna's price by 80 percent, calling it the company's "fastest and most affordable model" in a move that sent ripples through the industry. They're not messing around. Anthropic fired back with Claude Opus 5, touting "frontier intelligence… at half the price" of its flagship Fable 5. And the numbers don't lie. The combined effect has been dramatic, as Silicon Data's token price index shows prices for leading US lab models have dropped nearly a quarter since mid-July, a stunning shift that no one predicted just weeks ago. It's a price war now.

Tokens are the basic units of data that language models process. They form the backbone of most customer billing, and for companies running AI at scale, even small per-token savings translate into substantial monthly reductions. So that's exactly why the price war matters. It matters to the businesses signing these contracts.

Open Models Close the Gap

The price pressure stems from a fundamental shift in the AI landscape. It's a real shake-up. US labs have long competed on raw performance, charging premium prices for proprietary "closed" models, and they've built their entire business strategy around that exclusivity. But increasingly capable "open" Chinese models, which developers can freely download and modify, have eroded that advantage. So the old edge is gone.

DoorDash and Airbnb have both confirmed they now use Chinese-made models to control costs. That's a fact. These aren't fringe players experimenting on the side, and they don't see this as a temporary hack, but rather as a core part of their long-term strategy for managing AI infrastructure across their massive platforms. So it's a serious shift, not a sideshow. But the real story here is that major platforms are making strategic decisions about their AI infrastructure, and cost control is the driving force behind that choice.

That shift has coincided with a wave of releases from Chinese labs that have narrowed the performance gap with leading US models. So the US tech industry is watching nervously, worried that American developers could lose customers even while spending heavily to maintain their technological edge. But they're not panicking. Not yet.

What the Price Cuts Actually Look Like

The specific numbers tell a clear story. OpenAI cut GPT-5.6 Luna from $1 to $0.20 per million input tokens, and from $6 to $1.20 per million output tokens. Anthropic launched Opus 5 at $5 per million input tokens and $25 per million output tokens, exactly half the price of Fable 5.

Anthropic called off its planned September price increase for Sonnet 5 this week. That's a big deal. But the reversal signals just how sensitive the market has become to pricing, so the company clearly can't afford to test customer loyalty with even a modest hike right now.

But headline token prices don't tell the whole story. They really don't. More capable models can sometimes complete tasks using fewer tokens or with fewer attempts, so a model that looks expensive on paper might actually cost less per completed task when you factor in the efficiency of its output. Most models also operate at different "effort" settings, which adjust the computing power used to answer a question and affect both performance and final cost. And that's where the real math gets tricky.

The Benchmark Reality Check

Artificial Analysis, which benchmarks models across math, science, coding, and reasoning, found some surprising equivalences. Anthropic's Opus 5 at "medium" effort delivered similar performance and cost per task to Moonshot's Kimi K3 at "max" effort. OpenAI's GPT-5.6 Luna at "max" effort performed similarly to DeepSeek's V4 Flash at "max," but cost just under twice as much per task.

Those comparisons blur the line between "premium" and "budget" AI, making it harder for US labs to justify their higher prices.

The IPO Pressure Behind the Cuts

The price war isn't happening in a vacuum. So both OpenAI and Anthropic are plotting initial public offerings at trillion-dollar valuations, a financial maneuver that would dwarf nearly every company on earth. Investors want evidence that the industry's massive spending on AI can actually generate returns, though. That's the real test.

Corporate AI users are feeling their own cost pressures. That's hitting hard. Both Anthropic and OpenAI have been shifting some enterprise customers away from flat subscriptions toward usage-based billing, where companies pay according to the computational resources they consume, and that shift is forcing finance teams to rethink their budgets almost overnight. So some businesses have responded by imposing caps on AI usage or testing cheaper alternatives. It's a scramble.

Mantas Lukauskas, AI tech lead at Hostinger, a website hosting provider that has used large language models since 2020, noted that prices for the very best models were "flat to rising." He added that the recent pricing changes are the "first real test" of whether groups such as Anthropic and OpenAI can protect the cost of their most advanced offerings: "The US labs have cut the middle and are defending the top."

That quote captures the strategic dilemma perfectly. The US labs are sacrificing their mid-tier products to compete with Chinese rivals, but they're trying to hold the line on their most advanced, most expensive offerings.

Silence from the Labs

Anthropic and OpenAI both declined to comment on the pricing moves. So a person close to Anthropic said Opus 5's pricing below its flagship Fable 5 was simply how the startup's "family of models is built, there's no connection to competitors." That's it. Plain and simple.

That explanation stretches credibility. The timing, the magnitude, and the competitive pressure all point to a coordinated response to the Chinese threat.

The price cuts reflect a structural shift in how AI companies actually make money. It's a simple truth. As usage-based billing becomes the norm, per-token pricing turns into the primary lever for attracting and retaining customers, and that reality now forces every player in the market to rethink their margins from the ground up. But the Chinese rivals have simply exposed how much room there was to cut. That's the whole story.

The coming months will reveal whether the US labs can hold their premium pricing at the top end while defending against the Chinese challenge at the middle. It's a real test. Hostinger's Lukauskas sees this as the industry's first real test of pricing power, a moment that will show if these companies can actually dictate terms when the pressure comes from both directions at once. But if the US labs can't protect their most advanced offerings, the trillion-dollar valuations start to look very shaky. They're built on a lot of hope.

The winners are clear for now. Businesses deploying AI at scale are pocketing better models at lower prices, while their Chinese rivals have proven they can match the world's best on both cost and performance, a double win that reshapes the competitive landscape overnight. So the US labs are left to defend their turf. But they can't give away the store.