Alibaba's latest AI model claims to match Anthropic's top system, fueling China's challenge to US dominance

Alibaba has launched Qwen3.8-Max, its most advanced AI system yet, asserting that it rivals the capabilities of OpenAI and Anthropic's flagship models. The company made the system broadly accessible to users this week, following a preview last month in which it positioned the model as second only to Anthropic's Claude Fable 5.

Performance claims backed by benchmarks#

Alibaba's own evaluations and rankings on Arena.AI, a crowdsourced model-comparison tool, support the company's performance assertions. Testing conducted by Alibaba indicates that Qwen3.8-Max matches—and occasionally surpasses—Fable 5 on standard benchmarks. On Arena's text-comparison leaderboard, the new model ranks behind only Fable 5 and three variants of Anthropic's Opus family. For code-writing tasks, it trails only two Opus models and a competing system from Moonshot AI, while in image analysis, only Fable 5 outperforms it.

The model contains 2.4 trillion parameters, a measure of the numerical settings that enable an AI system to process information and recognize patterns. Though parameter counts serve as a rough proxy for capability, the relationship is not straightforward—larger does not always equal superior. Moonshot's competing model, Kimi K3, contains 2.8 trillion parameters, while leading US laboratories typically keep their exact figures confidential.

Open weights strategy as competitive lever#

Alibaba intends to release Qwen3.8-Max's weights—the adjustable values that govern how the system processes inputs—within the coming week. Unlike traditional proprietary AI, open-weight models grant developers substantial control, though not the full transparency of open-source software. This move marks a return to openness for Alibaba, which had shifted toward keeping its strongest models proprietary earlier in 2026.

Open-weight releases have emerged as a defining feature of China's AI landscape, with major domestic firms routinely making model weights publicly available. Moonshot released Kimi K3's weights last week, and this strategy has become widespread across Chinese AI companies. Beijing has promoted the practice as a way to expand China's standing in global AI governance and drive worldwide adoption of domestic alternatives.

Escalating competition and geopolitical stakes#

Alibaba's announcement intensifies the competitive push from Chinese firms, which appear to be steadily closing the performance gap with American counterparts. Multiple capable models have rolled out in recent weeks: Qwen3.8-Max follows closely on Moonshot's K3 launch, and both ByteDance and MiniMax unveiled advanced video-generation systems on Friday.