Understand this story
Qwen released the Qwen4-Max preview version on November 11, 2026, and announced that Qwen4 will be officially released soon. According to a November 13 report by 36Kr·AI, Alibaba-affiliated products have already integrated the preview version, while the official version has not yet launched. The report used the same prompt to have the Qwen4-Max preview version and Tencent's Hy4 each create a version of a webpage and a 3D animation, and concluded that Hy4 was richer in visual completeness and interaction details, while the Qwen4 version's visuals were more rough and restrained. The report also mentioned that Qwen-4-Flash-Early scored 49.9 points on the internal beta leaderboard GameFeel Bench, but noted that the leaderboard has only 12 stars and that 4 of its 7 metrics rely on manual scoring, so it should be viewed with caution. There are currently no subsequent developments.
AI-generated from public reports
Latest update · Discussion
Matthew Berman joined this discussionDevelopments
2 confirmed developments
- Source:Qwen Blog
- Source:36Kr · AI
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 16
Matthew BermanRelated discussion
Matthew Berman joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
ArenaRelated discussion
Arena joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
UnslothFirst-party
Qwen releases Qwen4-Max and previews that smaller Qwen4 models will be open-sourced
Qwen releases Qwen4-Max, built on a new MoE architecture, with native vision support and a 1M-token context window; the API is already live on QwenCloud. The company says that in its internal evaluations, compared with Qwen3.8-Max, it uses up to about 50% fewer tokens, and the average cost per task drops by more than 60%; input is $1.5/M tokens, output is $4.5/M tokens, and implicit cache is $0.075/M tokens. Qwen also said it will open-source smaller Qwen4 models, and Unsloth relayed that Qwen4-27B can run in a 16GB memory environment.
TeortaxesRelated discussion
Teortaxes joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Tongyi Qianwen / QwenFirst-partySelected
Alibaba's Tongyi Qianwen releases the Qwen4-Max model
Alibaba's Tongyi Qianwen has released Qwen4-Max, which uses a new MoE architecture, natively supports vision, and has a 1M token context window. The company says that in internal evaluations it surpasses Qwen3.8-Max using up to about 50% fewer tokens, with average per-task cost dropping by more than 60%; it scores 52.69 on the Work360 evaluation, about 10% higher than Qwen3.8-Max. The API is already live on QwenCloud, at $1.5/M tokens for input, $4.5/M tokens for output, and $0.075/M tokens for implicit caching, and smaller Qwen4 model weights will be open-sourced later.
Tongyi Qianwen / QwenFirst-partySelected
Alibaba Releases Qwen4-Max: MoE Architecture, 1 Million Token Context, Per-Task Cost Reduced to $1.67
Alibaba's Tongyi Qianwen has released Qwen4-Max, which adopts a new MoE architecture, natively supports vision, and has a 1 million token context window. According to the official announcement, on Work360 token usage is reduced by about 24% (Xhigh) and 35% (High) compared with Qwen3.8-Max, and on Terminal-Bench 4.0 by about 48% and 49%, while the average per-task cost on Work360 drops from $4.43 to $1.67. The API is now live on QwenCloud, priced at $1.5 per million input tokens, $4.5 per million output tokens, and $0.075 per million tokens for implicit caching; it can also be tried in Qwen Studio, QwenWork, and Qoder, and smaller Qwen4 model weights will be open-sourced later.
Qwen BlogFirst-party
Alibaba Qwen releases Qwen4-Max: 2.4T parameters, one-million-token context, focused on long-horizon agents
Alibaba Qwen has released Qwen4-Max, built on the new Qwen4 architecture, featuring a 2.4-trillion-parameter backbone, native vision capabilities, and a one-million-token context window. According to the official announcement, its score on the Work360 benchmark rose from 47.74 for Qwen3.8-Max to 52.69, while the average cost per Work360 task dropped from $4.43 to $1.67, a decrease of 62%. The model is now available on QwenCloud and can be used via Token Plan or API; it is also integrated into QwenWork, Qoder, and QwenStudio. The company says smaller open-source models in the Qwen4 series will be released soon.
Nov 15
Alibaba Cloud Developer Community AI
Alibaba Cloud Model Studio (Bailian) Large Model Platform Double 11 Buying Guide: Six Modules, Model Matrix, and Pricing Strategy
Alibaba Cloud Model Studio (Bailian) large model service platform is offering a 50% discount for new customers on Token Plans during Double 11 2026 (November 1 to 30). The Lite tier costs 19.5 yuan/month and includes 11,500 Credits, while the Pro tier costs 249.5 yuan/month and includes 180,000 Credits. The platform brings together about 190 models; the flagship Qwen3.8-Max is a 2.4-trillion-parameter MoE model with a 1-million-token context window, priced at 12 yuan/36 yuan per million tokens for input/output. The Team edition is not eligible for the 50% discount; a standard seat costs 150 yuan/month and includes 25,000 Credits.
Nov 13
36Kr · AI
Qwen4-Max Preview Released, Compared Alongside Tencent Hy4 in the Same Period
Qwen released the Qwen4-Max preview on November 11 and announced that Qwen4 will be officially released soon; Alibaba-affiliated products have already integrated it, while the official version has not yet launched. The author used the same prompt to have the Qwen4-Max preview and Tencent Hy4 each produce a version of a web page and a 3D animation, concluding that Hy4 is richer in visual completeness and interaction detail, while the Qwen4 version's visuals converge rather coarsely. The article also mentions that on the GameFeel Bench internal beta leaderboard, Qwen-4-Flash-Early scored 49.9 points, but the leaderboard has only 12 stars and 4 of its 7 metrics rely on manual scoring, so it should be viewed with caution.