Alibaba Officially Releases Qwen4, Advancing Agent and Coding Capabilities Toward Real-World Full Scenarios
On November 16, Alibaba released its new-generation foundation large model Qwen4-Max
On November 16, Alibaba released its new-generation foundation large model Qwen4-Max, with 2.4T total parameters, native vision capabilities, and support for million-token context. Based on an innovative model architecture and training optimizations, Qwen4-Max achieves higher performance with lower token consumption, and continues to evolve in key capability areas such as agents and coding: in the authoritative global front-end coding blind test CodeArena, Qwen4 ranks sixth; in third-party evaluations of agent capabilities and professional tasks such as Agents' Last Exam (ALE) and Vals Index, Qwen4 is in the leading tier. Qwen4 is deeply optimized for real tasks across all domains, can better invoke professional tools to handle long-cycle tasks, and advances the deep deployment of AI in professional fields such as automotive manufacturing, game development, financial investment research, and scientific research.
Qwen4-Max is now available on the Qwen AI platform to provide API services externally, with cache hits costing as low as 0.5 yuan per million tokens, and the average cost of completing a single task dropped by 60% compared with the previous generation model; users can also call the new model through the Alibaba Token Plan, or experience it directly in applications such as Qwen Office, Qoder, and the Qwen APP. It is reported that the small and medium-sized models in the Qwen4 series will be open-sourced and released successively.

(Caption: In the CodeArena global front-end coding leaderboard, Qwen4-Max ranks 6th)
Qwen4-Max adopts a new MoE architecture based on Qwen Sparse Attention (QSA), with 2.4T total parameters activating only 71B, 25% fewer activated parameters per token than the previous generation model, and higher thinking efficiency: in the Terminal-Bench 4.0 professional coding evaluation, at the same Xhigh thinking level, Qwen4-Max's average token usage per attempt dropped from 289K to 151K compared with the previous generation Qwen3.8-Max, a reduction of nearly 50%, while the evaluation score rose from 38.9 to 48.5, an increase of nearly 25%. At the same time, thanks to the sharp drop in cache-hit pricing, evaluation results show that Qwen4-Max's average Cost per Task compared with Qwen3.8-Max decreased by 58%.
The brand-new architecture combined with efficient training optimizations has comprehensively improved Qwen4-Max's performance. In the CodeArena global blind-test authoritative leaderboard focused on coding, Qwen4-Max ranks 6th globally in front-end (Webdev), sitting between GPT6-Astra and Claude Fable5.1, setting a new best result for Chinese models; in the "Agents' Last Exam" (ALE), it achieved the highest score of 46.7, a substantial increase of 18.8 points over Qwen3.8-Max; in the Chartography evaluation, which assesses large models' ability to understand professional charts and reason, Qwen4-Max scored 54.2, an improvement of 12.2 points over the previous generation model; in the professional work evaluation JobBench, the new model scored 68.5, surpassing GPT-6.1sol and trailing Claude Opus5.5; in the HLAB-Vals-1.1 evaluation focused on the legal profession, Qwen4-Max scored 21.3, a gain of 5.9 points over Qwen3.8-Max, placing it among the industry's leading models.
Whether for simple everyday needs or long-cycle complex tasks, Qwen4 can deliver reliably, making human-machine collaboration smoother and more efficient. Qwen4-Max implements a Live Delivery mechanism within the model, which can first generate a usable initial version and continuously expand the content version by version according to user feedback until completion: for a design request for a physics classroom popular science website, Qwen4-Max delivered 14 usable versions, ultimately expanding into a refined website containing 3D animations, interactive experiments, and detailed principle explanations. At the same time, Qwen4-Max also possesses the ability to perform operations research modeling, solving, and verification around complex objectives and constraints, and through autonomous planning and continuous iteration, orchestrates agents to solve more complex tasks: based on Alibaba Cloud AI infrastructure, Qwen4-Max autonomously and closed-loop optimized the operating systems supporting its own applications, inference, and training, reducing model cost expenses by 10% in data annotation scenarios, and increasing the optimal solving speed in CPU offline rescheduling scenarios by more than 14 times.

Caption: Across 721 directions of real work domains in full scenarios, Qwen4 improves overall professional capability by more than 10% compared with the previous generation model
From professional collaboration (Cowork) to all-weather agents, Qwen4 is expanding deeper into more real work tasks and production environments. The Qwen team built the Work360 full-scenario modeling system covering 721 subdomains, and in systematic evaluations, Qwen4-Max scored 52.69, an improvement of more than 10% over the previous generation model. In professional design fields such as game development, Qwen4 demonstrates strong 3D modeling capabilities and can better invoke professional tools such as Blender, generating a fun and cool 3D game with a single sentence. In the automotive manufacturing industry, Qwen4 has run through the entire process of production line planning, part design, and machining verification, drastically reducing the manual time for machining programming of a single workpiece from 475 minutes to 24 minutes, a reduction of 95%; in financial investment research scenarios, Qwen4 proficiently invokes 90 MCP professional tools, produces Excel workpapers and PPT reports, and supports dual traceability to ensure reliable conclusions.
In the frontier field of AI-assisted scientific discovery (AI For Science), Qwen4 has initially produced real results in materials and drug discovery. In the field of materials discovery, Qwen4 operated autonomously for 48 consecutive hours, selected candidate molecules from 2.7 million candidate molecules, and through 4 rounds of the "design—synthesis—test" closed loop, actually synthesized a new type of coolant material, providing a new idea for alleviating the data center cooling challenges brought by high-power chips; in the field of drug discovery, targeting a key target for Alzheimer's disease, Qwen4 completed 3,239 computational evaluations through 454 tool calls, designed and optimized 24 candidate molecules, and the nominal binding affinity of the finally selected molecule to the target improved by about 3.6 times compared with the starting molecule, making it a more effective new option.
This article is provided by Alibaba; QbitAI is authorized to repost it, and the views belong to the original author.

- Deadline today! 2026 AI Awards, miss it and wait another year! 2026-11-16
- MEET2027's value rises further: AI4S/Quantum Computing/AI Infra… The second batch of guests is coming in strong! 2026-11-13
- BYD's 19 millionth new energy vehicle rolls off the line, tech new luxury flagship SUV Fangchengbao Tai 9 opens pre-orders 2026-11-13
- From Yangwang's new top-speed record and Dingyao's global launch, looking at BYD's path of internationalization and premiumization 2026-11-13
Related Reading

Alibaba releases the "2026 AI for Good Action Report," sustainable development of AI data centers becomes the annual highlight
On November 5, Alibaba released the "2026 AI for Good Action Report"

Alibaba DAMO Academy's Qingcheng Award presented again! 10 mainland young scientists each receive 1 million, youngest winner is 28
This year is the second edition, with two women selected.
Alibaba became a sucker?? An open-source project acquired for $100 million, the core team left to build a competitor, then sold it for another $100 million

Alibaba DAMO Academy scientist built Xiaomandong in 3 years! Mass-produced logistics robot, fully autonomous driving, 100 kilometers on 4 kWh of electricity
The culmination of Alibaba DAMO Academy's most cutting-edge AI and autonomous driving technologies

Alibaba announces new progress on omnimodal models, Qwen4 and next-generation video model both in training
On September 22, the 2026 Apsara Conference opened, and Alibaba announced the latest progress on large models.
