AB
AB
AiBoss

All Categories

project

Qwen3-Max-Thinking - Alibaba's flagship reasoning model

Qwen3-Max-Thinking is Alibaba's latest flagship inference model for the Thousand Questions platform, boasting over 1 trillion parameters and 36TB of pre-trained tokens. Through test-time scaling and reinforcement learning, the model significantly improves inference performance and efficiency...

What is Qwen3-Max-Thinking?

Qwen3-Max-Thinking is Alibaba's latest flagship inference model for Qianwen, boasting over 1 trillion parameters and 36TB of pre-trained tokens. Through test-time scaling and reinforcement learning, the model significantly improves inference performance and efficiency. In 19 major benchmark tests, it matches or even surpasses top closed-source models such as GPT-5.2-Thinking, Claude-Opus-4.5, and Gemini 3 Pro. The model greatly enhances native agent capabilities, enabling it to autonomously invoke tools and provide more intelligent and fluent answers. Developers and general users can experience it for free through Qwen Chat, the Qianwen PC platform, and the web interface. Enterprises can obtain API services (qwen3-max-2026-01-23) through Alibaba Cloud's Bailian platform, and the new model will also be integrated into the Qianwen app.

Main functions of Qwen3-Max-Thinking

  • Strong reasoning abilityIt excels in complex reasoning, mathematical calculation, and logical analysis, and can handle highly challenging reasoning tasks, such as math competition problems and complex logic problems.
  • Autonomous tool callIt possesses powerful agent capabilities, enabling it to autonomously invoke tools such as search engines and code interpreters to provide more accurate and intelligent answers.
  • Efficient reasoning mechanismBy using test-time scaling technology, inference efficiency is improved, redundant calculations are reduced, and more efficient inference results are achieved.
  • Multi-domain applicationsIt is applicable to a variety of scenarios, including code generation, data analysis, and content creation, and can provide powerful technical support for developers and users.

The technical principles of Qwen3-Max-Thinking

  • Ultra-large-scale parameters and dataThe model has over 1 trillion parameters and 36T of pre-trained data, which enhances the model's knowledge reserves and reasoning ability through training with massive amounts of data.
  • Expansion mechanism during testingDuring the reasoning phase, an experience-based, multi-round iterative expansion strategy is adopted to avoid redundant reasoning, focus on unresolved uncertainties, and significantly improve reasoning efficiency and performance.
  • Adaptive tool callsThrough a specially designed training process, the model can autonomously select and call built-in tools (such as search, code interpreter, etc.) according to task requirements, achieving more intelligent interaction and problem solving.
  • Reinforcement learning optimization: Conduct joint reinforcement learning training based on rules and model rewards on diverse tasks to further improve the model's performance and ability to align with human preferences.
  • High-efficiency inference architectureBy optimizing the allocation of computing resources and inference paths, computational redundancy is reduced, and overall inference efficiency is improved, enabling it to perform well in complex tasks.

Performance of Qwen3-Max-Thinking

  • High-difficulty task handlingIn multiple authoritative benchmark tests, Qwen3-Max-Thinking has set new world-leading (SOTA) performance records. For example, in the tool-enabled "Last Test of Man" (HLE), Qwen3-Max-Thinking scored 58.3, significantly surpassing GPT-5.2-Thinking (45.5) and Gemini 3 Pro (45.8), becoming the highest score among all current models.
  • Mathematical reasoningIn the IMO-AnswerBench, a mathematical ability test benchmark at the IMO difficulty level, Qwen3-Max-Thinking achieved the highest score of 91.5, demonstrating strong mathematical reasoning ability.

How to use Qwen3-Max-Thinking

  • Developers can try Qwen3-Max-Thinking for free on Qwen Chat.
  • Enterprises obtain API services through Alibaba Cloud's Bailian platform.
  • Regular users can try it out on the PC and web versions of Qianwen. The Qianwen APP will be available soon, and all users can experience it for free.

Application scenarios of Qwen3-Max-Thinking

  • Solutions to complex problemsQwen3-Max-Thinking can efficiently solve challenging mathematical, logical, and scientific problems, providing accurate reasoning and solutions.
  • Code generation and optimizationThe model supports the rapid generation of high-quality code, optimizes the programming process, and improves development efficiency.
  • Data Analysis and Report GenerationThe model can generate professional reports, provide data support for decision-making, and help companies gain insights into market dynamics.
  • Content creationThe model can help users write articles, creative copy, and stories, improving content creation efficiency.
  • Educational guidanceIt provides students with answers to academic questions and personalized learning suggestions to support teaching and improve learning outcomes.
Back