Alibaba Prepares 10-Trillion-Parameter Model
Alibaba announced plans to scale Qwen models to 10 trillion parameters and unveiled the new Zhenwu V900 AI chip.
Alibaba Prepares 10-Trillion-Parameter Model
The company shared its plans for the next generations of models: Qwen 4 is currently in training, while Qwen 4.5 and Qwen 5 are planned to be scaled to 5–10 trillion parameters. For comparison, the current Qwen 3.8 Max contains 2.4 trillion parameters.
In parallel, Alibaba introduced a new AI chip, Zhenwu V900, which is three times more powerful than the previous generation. Based on this chip, the company plans to build clusters of up to 500,000 GPUs and reach a data center power capacity of 20 GW by 2032.
[.reuters.com/business/retail-consumer/alibaba-plans-ai-model-with-5-trillion-10-trillion-parameters-unveils-new-chip-2026-09-22/)
Why it matters
AnalysisScaling parameters and developing proprietary chips demonstrate Alibaba's ambition for full independence in AI infrastructure and leadership in the large language model market.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Part II. AI Agent Architecture: From Model to Memory and Tools
A deep engineering dive: tokenization, KV-cache, sampling determinism, the training stack, quantization, context engineering, the safety stack, production RAG, ReAct, and multi-agent swarm systems.
Frontier Models
The most powerful class of artificial intelligence at the forefront of global research (Claude 3.7 Sonnet, OpenAI o3/GPT-4.5, Gemini 2.0 Pro), defining the limits of modern reasoning, autonomy, and coding capabilities.
LLM (Large Language Model)
A fundamental class of neural network architectures based on autoregressive transformers, predicting the probabilistic distribution of subsequent tokens and demonstrating emergent properties of abstract reasoning, code synthesis, and logical inference.
Thinking Machines Releases Inkling 2 Small: a 594B MoE Model That Can Fine-Tune Other Models
Thinking Machines released Inkling 2 Small, a 594B-parameter MoE model that leads among open models in several benchmarks and demonstrated autonomous fine-tuning of Qwen3.5-4B.