Alibaba Group Holding Ltd. released Qwen3.8-Max this week, the company’s most advanced model to date, and the headline number did most of the talking. The new open-weight release appeared to match or exceed Anthropic’s flagship Fable 5 on the same benchmarks, and it landed at a price point that makes the comparison almost unfair. Five Chinese frontier models in eight weeks have moved the question from whether China can build frontier AI to whether Silicon Valley can keep charging for it.
The Los Angeles Times laid out the cascade on Tuesday. Qwen3.8-Max followed Moonshot AI’s Kimi K3, which two weeks earlier showed performance comparable to the priciest U.S. options on a far smaller budget. ByteDance’s Seedance, version 2.5, dominates video generation after OpenAI shelved Sora. DeepSeek returned this summer with V4 Flash, a model that prices a complex real-world workload at $0.03 against Claude Fable 5’s $3.15, roughly a hundred-fold gap. Z.ai’s GLM-5.2, released in June, was the highest-ranked open-source model globally at launch.
Death Zone Economics
The benchmark chart from independent evaluator Artificial Analysis now shows a region labeled the DeepSeek death zone. Any model that charges more for the same product or matches the same price with worse capability effectively loses the contest. Mid-market rivals are now under pressure to slash prices to match the Hangzhou startup or spend heavily to build smarter models. GLM-5.2, Kimi K3, and Qwen3.8-Max all sit above the line in a higher performance bracket.
The economics for U.S. labs are eroding at exactly the moment their business plans depend on defending premium pricing. OpenAI and Anthropic are both planning initial public offerings at valuations north of a trillion dollars, a mark that assumes high-margin per-token revenue for the foreseeable future. Both labs also keep parameter counts confidential, unlike Chinese competitors that publish specific figures. Moonshot and Alibaba have built very large models, each with more than two trillion parameters, and then moderated compute costs by activating only segments at a time.
Where Pricing Power Goes Next
Internal Chinese competition has accelerated timelines as much as the global race has. DeepSeek disrupted the Chinese contest as much as it did the international one. Some Chinese companies are sacrificing near-term profit for users and developer mindshare. China’s AI providers have become trapped in a brutal price war, prioritizing market share over profitability, according to Bloomberg Intelligence senior analyst Rob Lea. ByteDance’s Seedance has pushed the model in video. The price competition has reached ninety-nine percent discounts to prevailing market rates.
U.S. enterprises are starting to architect around the price gap. Dermot McGrath, founder of Shanghai-based startup consultancy ZenGen Labs, said he still uses Anthropic’s Claude Code as an architect to draft an action plan, then hands execution off to DeepSeek within the same environment. A few months ago I wouldn’t have done that, McGrath said. The Chinese models weren’t as reliable at tool calling or long-running agent workflows. Alibaba made improvements in long-horizon execution with its Qwen release. Beijing-based Z.ai also made long-horizon performance a priority.
The Washington Response
The White House is watching. Both President Donald Trump and his counterpart Xi Jinping have made AI leadership a national priority. When Xi visits the White House in September, because of Kimi, because of Alibaba and the model they just released last night, Xi will definitely feel like he has, in Trump’s language, more cards to play, said George Chen, partner and Digital Practice chair at The Asia Group. Trump told reporters in the Oval Office last Wednesday that he was weighing the China threat against the need for additional safety controls. We have to be careful in both ways. We don’t want to restrict them where all of a sudden, we come in second to China.
Washington has already moved on the IP question. The administration threatened sanctions over intellectual property concerns after Moonshot’s Kimi breakthrough, and earlier instructed Anthropic to withdraw its Fable Mythos-class model and OpenAI to limit access to GPT-5.6, actions that have heightened enterprise interest in models less vulnerable to policy-driven access restrictions. If there weren’t these Chinese open-source models, OpenAI and Anthropic would be laughing all the way to the bank, said Kai-Fu Lee, whose startup 01.AI offers open models to global clients. Now there’s an alternative, and it’s cheaper.
Analyst quote
“The most important change since January 2025 is that China’s progress no longer looks like a single-company breakthrough. The first DeepSeek moment looked exceptional. The recent releases suggest China now has a repeatable system for producing models close to the global frontier.”
Poe Zhao, founder of the Hello China Tech newsletter, quoted by the Los Angeles Times
What Comes Next
The U.S. response has split between frontier-model restraint and a race-to-the-bottom in the long tail. Enterprise buyers are not waiting for a policy resolution. The cost of AI work has inflated over recent months, McGrath said, and efficiency has risen to be an even greater priority. The first IPO prospectus to disclose that more than a quarter of inference volume is being routed through Chinese open-weight models will reset the entire pricing conversation. Until then, the death zone is expanding across the chart.