Alibaba Releases Qwen3.6-Plus With 1M Token Context — Proprietary Shift Signals $100B AI Revenue Target

On April 2, 2026, Alibaba released Qwen3.6-Plus via Alibaba Cloud Model Studio API, marking its third proprietary AI model release in just a few days. The model features a one-million-token context window and is specifically optimized for agentic coding tasks including frontend development and complex code generation.

According to Alibaba's benchmarks, Qwen3.6-Plus partially outperforms Anthropic's Claude 4.5 Opus (the older flagship model replaced by 4.6 Opus in December 2025). However, Claude 4.6 Opus still leads on some benchmarks — scoring 65.4% on Terminal-Bench 2.0 versus Qwen3.6-Plus. The Qwen team claims significant improvements in agentic coding capabilities over the previous Qwen3.5 generation.

This release represents a major strategic pivot for Alibaba. The company had previously released Qwen models as open source, building a massive community and adoption base. The shift to proprietary models — including the recent Qwen3.5-Omni which is also not freely available — signals Alibaba's intention to monetize its AI investments through enterprise customers.

According to Bloomberg, Alibaba is targeting $100 billion in AI revenue over the next five years. The company's cloud division faces intense competition from ByteDance, and proprietary models are seen as key to driving enterprise revenue.

Qwen3.6-Plus will be integrated into the Qwen chatbot app and Alibaba's new enterprise AI service called Wukong. The 1M context window positions it for enterprise use cases requiring long-document processing, code repository analysis, and extended agentic workflows.

The release came on the same day as Google's Gemma 4, creating a fascinating contrast: Google moving FROM proprietary TO open (Apache 2.0), while Alibaba moves FROM open TO proprietary. This divergence reflects fundamentally different business strategies in the AI model market.

The model was trending on HackerNews with 322 points and 109 comments, with developers noting the impressive agentic coding capabilities but expressing disappointment about the proprietary shift.