Alibaba releases Qwen4-Max, API-hosted only and not open source
Understand this story
Alibaba's Qwen team releases Qwen4-Max, the first model in the Qwen4 series, featuring a 2.4T-parameter sparse MoE backbone with 71B activated per token, supporting text, image, and video input and a 1M-token context window. The model does not open-source its weights and cannot be self-deployed or used in quantized versions; it is available only through the QwenCloud API and hosted channels such as QwenWork, Qoder, and QwenStudio. Subscription access goes through the Token Plan, and what is opening this month is a Token Plan-only preview with a default rate limit of 5 million tokens per minute. According to The Decoder, Alibaba plans to open-source smaller Qwen4 models later. Calling Qwen4-Max through its own API costs $1.50 per million input tokens and $4.50 per million output tokens, with a cache-hit price of $0.075. The report positions the model as a competitor to Opus 5.5, GPT-6, and Kimi K3.1 Pro. Qwen4-Max itself still has no open weights, and a self-deployable open-source version has yet to arrive.
AI overview based on source reports · Updated 1 days ago
Coverage timeline
Follow the coverage to explore different perspectives.
- The DecoderFeaturedAlibaba's Qwen4-Max takes on Opus 5.5, GPT-6, and Kimi K3.1 Pro at much lower prices
Alibaba released Qwen4-Max, a 2.4-trillion-parameter mixture-of-experts model that activates 71 billion parameters per token, processes images natively, and offers a one-million-token context window using a hybrid attention architecture. Alibaba reports it beats Qwen3.8-Max on its internal evaluations while cutting average cost per Work360 task 62 percent to $1.67, and Artificial Analysis puts it at 43 points on its Intelligence Index at $3.27 per task, three behind Kimi K3.1 Pro. It is available through QwenCloud and in QwenWork, Qoder, and QwenStudio, with smaller Qwen4 models planned for open-source release.
- MarkTechPostFeaturedAlibaba Releases Qwen4-Max: 2.4T MoE, 71B Active, 1M Context
Alibaba's Qwen team released Qwen4-Max, a proprietary sparse MoE model with a 2.4T-parameter backbone, 71B active parameters per token, text/image/video input and a 1M-token context, available only via QwenCloud APIs with no open weights. Alibaba reports it ranks first on 4 of 16 charted benchmarks, led by HLAB-Vals-1.1 at 21.3, while trailing on coding and vision, and says average cost per task fell 62% with list prices of $1.50/$4.50 per million input/output tokens.
Activity for this story
- Current comparable activity
- 7
- Peak comparable activity
- 18Nov 17 01:00
- Change in the past 24 hours
- -50%
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.