Technology

Add as a preferred source on Google

Alibaba Says Qwen4-Max Outscored Claude and Kimi in Coding Tests

Alibaba kept the flagship’s weights closed and said smaller Qwen4 models would arrive soon.

Updated 1 min read
Employees walk past Alibaba’s Xixi campus in Hangzhou, China, beneath an overcast morning sky. / EconoTimes.ai (mono)
Employees walk past Alibaba’s Xixi campus in Hangzhou, China, beneath an overcast morning sky. / EconoTimes.ai (mono)

Alibaba says Qwen4-Max beat two named competitors in coding tests, adding a new model to competition with U.S. AI companies while keeping its flagship model unavailable as open weight.

The company said Qwen4-Max scored near the top of coding benchmarks and outperformed Anthropic’s Claude Fable 5.1 and Moonshot, a Chinese AI firm, with its Kimi K3.1 Pro model.

Alibaba did not release Qwen4-Max as an open-weight system, which lets users download and customize a model’s weights. It said smaller Qwen4 models would be available soon but gave no date. The launch follows an earlier Token Plan preview.

Bloomberg reported that Qwen had more downloads than other open-weight model families to date. The outlet also cited recent Chinese AI releases from Tencent, including Hunyuan 4, and Moonshot’s Kimi K3.1 Pro in describing competition with U.S. AI leaders.

Alibaba’s push also covers chips and computing capacity, Bloomberg reported. Alibaba CEO Eddie Wu said in September that its T-Head chipmaking unit would bring China’s most powerful AI chips to the domestic market. Wu also set a goal of 20 gigawatts of data center capacity worldwide by 2032.

Bloomberg reported that Tencent and Moonshot had also adjusted model safeguards. Alibaba added extra protections to Qwen4-Max amid concerns about cybersecurity risks from open-weight models.

Related articles

View all