Alibaba Launches Qwen4-Max, Its New Flagship AI Model Aimed at Anthropic and OpenAI

Alibaba has launched Qwen4-Max, the first model in its new Qwen4 family, claiming coding performance that beats Anthropic's Claude Fable 5.1 and Moonshot's Kimi K3.1 Pro. Unlike earlier Qwen models, the flagship is closed-weight and available only as a hosted API on Alibaba Cloud.

Nov 16, 2026 - 19:35
 0  7
Alibaba Launches Qwen4-Max, Its New Flagship AI Model Aimed at Anthropic and OpenAI
A smartphone displays the purple Qwen logo and wordmark, with a large blurred Qwen logo in the background.

Alibaba Launches Qwen4-Max, Its New Flagship AI Model Aimed at Anthropic and OpenAI

Alibaba has launched Qwen4-Max, the first model in its new Qwen4 family, pitching it as a direct rival to the leading American AI models. The Chinese company says the flagship scored near the top of coding benchmarks, ahead of Anthropic's Claude Fable 5.1 and homegrown rival Moonshot's Kimi K3.1 Pro, Bloomberg reported on Monday, November 16.

The launch is Alibaba's latest challenge to US leadership in artificial intelligence, and it arrives in the middle of a busy month for Chinese AI labs. Tencent has released Hunyuan 4, Moonshot has shipped Kimi K3.1 Pro, and Meituan recently unveiled a trillion-parameter model trained without Nvidia chips. Alibaba is now answering with its most capable model yet — but with a notable change of approach: Qwen4-Max is not open-weight.

From preview to full release in five days

Qwen4-Max first appeared on November 11 as a preview available only to subscribers of Alibaba's Token Plan, alongside early access in the company's own apps, including QwenWork, Qoder and the Qwen app. The full release followed on November 16, when Alibaba Cloud made the model generally available on its Model Studio platform with a public pay-as-you-go API.

The preview version, qwen4-max-preview, is scheduled to run until November 25 and will then be retired, according to Alibaba's documentation. The full model carries the ID qwen4-max.

Qwen4-Max succeeds Qwen3.8-Max, which arrived in August 2026. Alibaba describes it as the next-generation flagship of the Qwen family, built for coding, office work, front-end design and long-running agentic tasks. It is a multimodal model that reads text, images and video, reasons by default, and can hold up to one million tokens of context in a single session.

Coding claims — and a mixed picture

The headline claim is about coding. Alibaba says Qwen4-Max scored near the top of coding benchmarks, outperforming Claude Fable 5.1 and Kimi K3.1 Pro. Those are the company's own results, reported by Bloomberg, and they have not yet been independently verified.

Alibaba's own published charts tell a more mixed story. On the DeepSWE v1.1 coding test, Qwen4-Max scored 75.7, narrowly ahead of Claude Opus 5.5 at 74.2. On Terminal-Bench 4.0, however, it scored 48.5, trailing Opus 5.5 at 59.6 and Kimi K3.1 at 67.2. Until independent testers run the model on their own workloads, the benchmark charts are a claim, not proof — and the comparison is against Claude Fable 5.1, not Anthropic's newest models.

Under the hood, Qwen4-Max is built on a sparse mixture-of-experts design with a backbone of 2.4 trillion parameters. Alibaba says the new architecture is designed for the whole Qwen4 series, and that smaller versions of the family will follow.

Cheaper to run, but closed

Price is where Alibaba makes its strongest case. In Singapore, Qwen4-Max costs $1.50 per million input tokens and $4.50 per million output tokens — 25 percent below the $2 and $6 list prices of Qwen3.8-Max. A one-million-token free quota is also available in Singapore for 90 days.

The catch is access. Unlike earlier Qwen models, Qwen4-Max cannot be downloaded or run on a developer's own hardware. It is available only as a hosted service through Alibaba Cloud Model Studio, either through the pay-as-you-go API or the Token Plan subscription, and through partners such as Vercel's AI Gateway, which added the model on November 15. The model is listed in six regions — Beijing, Singapore, Frankfurt, Virginia, Tokyo and Hong Kong — while fine-tuning and custom deployment are not supported.

Alibaba says smaller Qwen4 models will be released openly, but it has given no date, sizes or licence terms. That is a break with the Qwen family's history: Alibaba's Qwen series is so far the most downloaded open-weight AI model family in the world, following an ambitious push into the technology spearheaded by chief executive Eddie Wu.

Why the launch matters

The release fits into a wider shift among Chinese AI labs. Alibaba introduced upgraded security measures in its latest model, following a series of alarming incidents that raised concerns about the cybersecurity risks of open-weight models. Tencent and Moonshot have taken similar steps with their recent releases.

For Alibaba, the model is one part of a much larger bet. The Hangzhou-headquartered company has made attaining artificial general intelligence its primary goal, and Wu has set a target of building 20 gigawatts of data centre capacity globally by 2032. Its T-Head chipmaking unit, which recently filed for an initial public offering in Hong Kong, is expected to roll out China's most powerful domestically made AI chips.

Whether Qwen4-Max lives up to its coding claims will become clearer as developers put it through real work. For now, Alibaba has delivered a flagship that is cheaper than its predecessor, closed to downloaders, and aimed squarely at the US leaders — a combination that says as much about where the AI race is heading as it does about the model itself.

Related coverage on Innopark IT: Meituan releases open weights for LongCat-2.5, a 1.6-trillion-parameter model trained without Nvidia chips and Export-banned Nvidia chips may have trained China's most powerful AI models, research finds.

Sources

  • The Straits Times / Bloomberg — Alibaba launches new flagship AI model in US riposte, November 16, 2026
  • The Latent Daily — Alibaba releases Qwen4-Max with 1M context and 25% lower prices, November 16, 2026
  • Unite.AI — Alibaba Cloud Releases Qwen4-Max with 25% Lower Singapore API Prices, November 16, 2026
  • Silicon Report — Alibaba launches closed-weight flagship AI model Qwen4-Max, November 16, 2026
  • The Model Atlas — Qwen4-Max is out: price, benchmarks and how to try it, November 16, 2026
  • APIDog — What Is Qwen4-Max? Alibaba's New Flagship Model, November 16, 2026

What's Your Reaction?

Like Like 0
Dislike Dislike 0
Love Love 0
Funny Funny 0
Angry Angry 0
Sad Sad 0
Wow Wow 0
Nahid Joy I’m Nahid Hasan Joy, a technology writer, web developer, and digital enthusiast with a strong interest in the ever-changing world of technology.