Alibaba Launches Qwen4-Max AI Model with Closed Weights
Alibaba has released Qwen4-Max, its new flagship AI model featuring 2.4 trillion parameters and multimodal capabilities. Unlike previous open-weight versions, this model keeps its weights closed but offers competitive coding scores and reduced costs. The launch highlights Alibaba's strategy to compete globally while retaining control over its core technology.
Key points
- Alibaba launched Qwen4-Max, a 2.4 trillion-parameter multimodal AI model available via API and apps.
- The model keeps its weights closed, marking a shift from earlier open-weight releases in the Qwen series.
- Alibaba claims Qwen4-Max leads in coding benchmarks, beating models like Claude Opus 5.5 on specific tests.
- The model reduces average task costs from $4.43 to $1.67 on internal Work360 tests.
- Pricing is set at $1.50 per million input tokens and $4.50 per million output tokens for developers.
Alibaba Enters Flagship AI Race with Closed Weights
Alibaba Group has officially launched Qwen4-Max, the latest flagship model in its Qwen family of artificial intelligence systems. Announced on Monday, November 16, through a blog post by the Qwen team, the model is positioned as an "AI teammate" designed to assist with work, creativity, and discovery. This release marks a significant strategic shift for the Chinese tech giant, as it opts to keep the model's weights closed rather than open-sourcing them, a practice that has characterized much of the Qwen series' public presence.
Technical Specifications and Multimodal Capabilities
Qwen4-Max is a mixture-of-experts model containing 2.4 trillion parameters. This architecture allows the system to activate only a portion of its parameters for each request, optimizing efficiency. The model is multimodal, meaning it can process and generate text, images, and video. It supports a context window of up to one million tokens per prompt, enabling it to handle extensive documents or long conversations. Alibaba states that the new model performs more work with fewer tokens compared to its predecessor, Qwen3.8-Max.
Performance Claims and Benchmark Discrepancies
In its launch materials, Alibaba claims that Qwen4-Max scores near the top of coding benchmarks. The company asserts that the model beats Anthropic’s Claude Opus 5.5 and Moonshot AI’s Kimi K3.1 Pro in specific tests. On the DeepSWE v1.1 coding test, Qwen4-Max scored 75.7, slightly ahead of Claude Opus 5.5, which scored 74.2. However, independent analysis by FourWeekMBA highlights mixed results in other areas. On the Terminal-Bench 4.0, Qwen4-Max scored 48.5, trailing behind Claude Opus 5.5 at 59.6 and Kimi K3.1 at 67.2. These scores are based on Alibaba’s internal testing and have not been independently verified by external sources.
Cost Reduction and Accessibility
A key selling point for Qwen4-Max is its cost efficiency. According to Alibaba’s internal Work360 test, the average cost of completing a task dropped from $4.43 with the previous model to $1.67 with Qwen4-Max. Developers can access the model through the QwenCloud API, priced at $1.50 per million input tokens and $4.50 per million output tokens. The model is also integrated into Alibaba’s Qwen Studio, QwenWork, and Qoder apps. A limited preview of the model began on November 11, preceding the full public announcement.
Safety and Strategic Implications
Alibaba emphasized safety in the launch, noting that it retained training data for defensive cybersecurity purposes. This move to close the weights may reflect a growing trend among major AI providers to protect proprietary data and model architectures as competition intensifies. By keeping Qwen4-Max closed, Alibaba aims to maintain a competitive edge in the global AI market while offering a powerful tool for enterprise and developer use. The launch underscores the rapid evolution of AI capabilities and the ongoing race to balance performance, cost, and security in large language models.
Key takeaways
- Alibaba launched Qwen4-Max, a 2.4 trillion-parameter multimodal AI model with closed weights.
- The model claims top-tier coding performance, though benchmarks show mixed results against competitors.
- Task costs decreased significantly from $4.43 to $1.67 on internal tests.
- Access is available via API at $1.50/$4.50 per million tokens and through Alibaba apps.
- The closed-weight strategy marks a departure from previous open-source releases in the Qwen series.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.