Alibaba Cloud (Qwen)

What it is. Alibaba Cloud’s Qwen division builds the Qwen family of foundation models — served through Alibaba Cloud Model Studio (Bailian) and the qwen.ai portal — plus open-weight releases on GitHub, Hugging Face and ModelScope. Why it matters. Qwen is the clearest case in China of a hyperscaler pairing frontier closed models (Qwen3.8-Max, Qwen3.8-Flash) with a genuinely open MoE release (Qwen3.8-2.4T-A95B), making it the reference point for both API and open-weight buyers. Key characteristics. A 2.4T-parameter MoE flagship with 1M-token context and multimodal input; six serving regions; the Qwen Code and Qoder agentic-coding products. What a professional should know. The open A95B release and the closed Max share an architecture but not a license — the open weight carries a custom MIT-style license with scale-triggered obligations for large operators, so read the license text before commercial deployment.

Qwen is the large language model family built by the Qwen team at Alibaba Group / Alibaba Cloud. The official portal is qwen.ai; models are served through Alibaba Cloud Model Studio (Bailian) in six regions (Beijing, Singapore, Hong Kong, Frankfurt, US Virginia, Tokyo), with open-weight releases on GitHub (QwenLM), Hugging Face (Qwen) and ModelScope.

The current flagship is Qwen3.8-Max (2.4T-parameter MoE, 1M-token context, multimodal input), whose open-weight base Qwen3.8-2.4T-A95B was released 2026-08-12. The Chinese name 千问 appears in official docs; the older “Tongyi Qianwen / 通义千问” branding was not confirmed on the fetched English pages.

Why it matters

China AI Hub analysis: Alibaba Cloud occupies a structurally unique position in the Chinese ecosystem — it is the only hyperscaler that combines frontier closed APIs, a first-party open MoE release, and a multi-region cloud distribution layer (Model Studio) under one roof, which makes Qwen the natural default for enterprises that want a single vendor across closed and open workloads. The database evidences this breadth directly: three foundation models, three agents (Qwen Code, Qwen-Agent, Qoder) and one API all resolve to the same company record. China AI Hub analysis indicates Qwen’s structural role is the integration anchor — its value to buyers is less any single model and more the consistency of model, tooling and cloud under one provider, with the open A95B release as the hedge against API lock-in. China AI Hub analysis indicates the open A95B release is the strategic hinge — it lets enterprises start on open weights and graduate to the managed cloud without leaving the Qwen stack, hedging the closed Max tier against lock-in. Qwen models are also orchestrated by third parties: Manus routes to Qwen per third-party descriptions, and the model-neutral platform Dify serves Qwen weights.

Entity hub

Models

• Qwen3.8-Max
• Qwen3.8-Flash
• Qwen3.8-2.4T-A95B
• Qwen3.7-Plus

Products

• Alibaba Cloud Model Studio (Bailian)

API

• Model Studio

Agents

• Qwen Code
• Qwen-Agent
• Qoder

Research / Technology

• Distillation
• Function Calling
• Mixture of Experts
• Multimodal AI
• RAG
• Tool Calling

Comparisons

• DeepSeek-V4-Pro vs Qwen3.8-Max
• Qwen3.8-Max vs GLM-5.3

Pricing

• Alibaba Cloud pricing

Benchmarks

• AutomationBench
• DeepSWE
• GPQA Diamond
• HLE
• SWE-bench
• Terminal-Bench

What is uncertain

• The older “Tongyi Qianwen / 通义千问” branding was not confirmed on the fetched English pages.
• The open A95B release carries a custom MIT-style license with scale-triggered obligations for large operators.

Sources

evidence_id	source_name	source_url	source_type	published	verified	confidence	conflict
src-companies-alibaba-cloud-1	Alibaba Cloud Model Studio — text generation model list	https://www.alibabacloud.com/help/en/model-studio/text-generation-model	Official documentation	—	2026-09-20	high	—
src-companies-alibaba-cloud-2	Qwen3.8 open-model repository README	https://github.com/QwenLM/Qwen3.8	Official documentation	—	2026-09-20	high	—
src-companies-alibaba-cloud-3	Hugging Face model card — Qwen3.8-2.4T-A95B	https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B	Model card	—	2026-09-20	high	—

Labels used above: Official fact (from Alibaba Cloud Model Studio and Qwen’s GitHub/Hugging Face sources), and China AI Hub analysis (our synthesis, always introduced as such).