Alibaba Unveils Qwen4-Max, Claims Performance Rivaling US Frontier Models
Alibaba Group released Qwen4-Max, a 2.4-trillion-parameter mixture-of-experts model with a 1-million-token context window, asserting performance that rivals Anthropic and OpenAI [2]. The launch marks the latest Chinese frontier model positioned as a challenger to US labs.
Technical specifications and availability
Qwen4-Max is not open-weight. According to a Bluesky post cited by Techmeme, the API is served from Singapore and priced at $1.50/$4.50 (per million input/output tokens, per industry convention) [2]. The company says the model beats its predecessor, Qwen3.8-Max, using up to roughly 50% fewer tokens for comparable results [2].
Alibaba also states it plans to release weights for smaller Qwen4 models, though no timeline was given [2].
Benchmark claims
Alibaba claims Qwen4-Max outperforms Fable 5.1 and Kimi K3.1 Pro on coding benchmarks. The company frames the result as evidence that its flagship now rivals Anthropic PBC and OpenAI, in what Techmeme describes as China's latest challenge to US leaders [2].
The reporting cited here includes only Alibaba's self-reported benchmarks. The benchmarks Alibaba cites compare Qwen4-Max to Kimi K3.1 Pro, a Chinese model from Moonshot AI, and Fable 5.1 — rather than to US frontier models such as those from Anthropic and OpenAI. Without third-party replication on standard suites, the parity claim remains unverified.
Context: US-China AI competition
The launch arrives amid US export controls that restrict China's access to foundry capacity [3]. Alibaba has invested heavily in domestic chip development through its T-Head unit, which filed for a Hong Kong IPO in September [3]. The company has pledged more than $53 billion over three years to expand its AI capabilities [3].
What remains unknown
The available reporting does not mention training compute, data sources, energy consumption, safety evaluations, or deployment details. Alibaba's "up to ~50% fewer tokens" claim does not specify the task distribution on which it was measured [2], and no timeline was given for open-weight smaller models [2].
Significance
If independently verified, Qwen4-Max would represent the closest a Chinese lab has come to matching US frontier capabilities on paper. Until then, it is a claim backed by self-reported benchmarks on selected rivals, priced competitively, and delivered from a jurisdiction that reflects the geopolitical contours of the current AI race.
Sources
- Technology | Bloomberg Professional Services (bloomberg.com)
- Techmeme (techmeme.com)
- Alibaba's T-Head files for Hong Kong IPO (briefs.co)