TL;DR
New Release: Alibaba unveiled Qwen 3.5 on February 16, an AI model engineered for autonomous task execution with visual agentic capabilities. Cost Efficiency: The model delivers 60% lower costs and eight times better performance on large workloads compared to its predecessor. Market Context: The launch follows ByteDance’s Doubao 2.0 release and precedes DeepSeek’s expected new model, intensifying competition in China’s AI market. Strategic Goal: Alibaba aims to compete against ByteDance’s market-leading Doubao chatbot and establish international enterprise adoption through aggressive open-source strategies.
Alibaba unveiled Qwen 3.5 on February 16 in China, an AI model engineered for the agentic AI era with the ability to execute complex tasks independently. The Chinese tech giant claims the model delivers 60% lower costs and eight times better performance on large workloads compared to its predecessor.
The launch positions Alibaba to compete more aggressively against U.S. rivals and domestic competitors, particularly ByteDance’s market-leading Doubao chatbot.
“Built for the agentic AI era, Qwen3.5 is designed to help developers and enterprises move faster and do more with the same compute, setting a new benchmark for capability per unit of inference cost”
Alibaba Performance and Cost Advantages
The technical specifications underscore Alibaba’s focus on enterprise adoption. Alibaba said Qwen 3.5 was 60% cheaper to use than its immediate predecessor, addressing enterprise concerns about AI infrastructure expenses.
The model delivers eight times better performance when processing large workloads, according to Alibaba’s internal benchmarks. Visual agentic capabilities distinguish Qwen 3.5 from prior versions.
The model can independently control mobile and desktop applications, interpreting visual interfaces to complete tasks without human intervention. This functionality allows the AI to navigate software environments, click buttons, and complete multi-step workflows across different platforms.
Alibaba claims the model outperforms major U.S. AI models on several benchmarks. Published results show Qwen 3.5 surpassing GPT-5.2, Claude Opus 4.5, and Gemini 3 Pro in selected tests. By enabling AI models to autonomously interact with existing software interfaces rather than requiring expensive custom integrations, Qwen 3.5 reduces both infrastructure costs and implementation complexity.
China’s Competitive AI Market
These performance claims emerge against a backdrop of intensifying domestic competition. ByteDance’s Doubao commands the largest user base among Chinese chatbot applications, approaching 200 million users.
The company released Doubao 2.0 on February 15, one day before Alibaba’s Qwen 3.5 announcement. The timing suggests intensifying competition for enterprise customers and developers.
Alibaba has struggled to match Doubao’s organic growth. The company ran a coupon campaign earlier this month that drove a seven-fold increase in active users. The day-before-day-after launch sequence between ByteDance and Alibaba reveals a defensive positioning battle, with Alibaba relying on promotional campaigns rather than organic adoption.
Prior Qwen Models
This latest release follows an accelerated development cadence that has compressed traditional AI model cycles. Alibaba’s successive model releases demonstrate how Chinese AI firms have compressed development cycles that previously took quarters into weeks.
The company released Qwen 2.5-Max in January 2025 as a direct response to DeepSeek’s viral breakthrough. The model claimed to “surpass GPT-4o, DeepSeek-V3, and Llama-3.1-405B in almost every aspect,” according to Alibaba’s Cloud Division. Independent benchmarks provided mixed results.
However, the competitive cycle shows no signs of slowing. DeepSeek is expected to release its new-generation model in the coming days, setting up another benchmark showdown between Chinese AI firms.
The competitive cycle has accelerated sharply since DeepSeek became the first Chinese AI company to achieve global recognition. Google DeepMind CEO Demis Hassabis noted that “China’s AI models may be just ‘months’ behind those developed in the US.” The assessment reflects narrowing technical gaps that have prompted U.S. policymakers to reconsider export controls on AI chips.
Beyond the domestic battleground, Chinese firms have pursued aggressive open-source strategies to expand their global footprint. Alex Lu, founder of LSY Consulting, explained that Chinese companies hope countries outside China will use these models to ensure applications are built on Chinese AI foundations. This creates market penetration opportunities while Alibaba competes domestically against ByteDance’s dominance.
“The hope is countries apart from China will use these models to ensure large amounts of applications are built on these Chinese models. That’s one way for Chinese companies to penetrate the market”
Alex Lu, founder of LSY Consulting (via CNBC)
Qwen 3.5’s combination of cost efficiency and agentic capabilities aligns with this strategy. With DeepSeek’s next model expected within days, the race to establish which Chinese AI models dominate international enterprise deployments enters a decisive phase.