TellWorldAI World Archive
Ask中文
News
Models
Organisations
People
Papers & concepts
Products & agents
Chips & robots
History
Topics
Ask中文

Home›Models

Model

Qwen3

Qwen3 is a family of large language models from Alibaba, released on 29 April 2025 under Apache 2.0. Its main feature is hybrid thinking, which lets one model reason step by step or answer directly.

Status AvailableModality TextOpen weights YesContext (K tokens) 128 Era Generative AI for everyone

Parameters
235B
License
Apache-2.0
Access
API, App, Download
Release date
Developer
Alibaba Cloud

What it is

Qwen3 is a family of large language models from Alibaba's Qwen team. It includes two mixture-of-experts (MoE) models and six dense models. Its main feature is hybrid thinking: one model can reason step by step or answer instantly, and the mode is switched with an enable_thinking option or with the /think and /no_think tags in a prompt.[1]

Origins

Qwen3 was released on 29 April 2025 under the Apache 2.0 license.[1][2]

Pre-training used about 36 trillion tokens, nearly double the 18 trillion used for Qwen2.5. The data covered 119 languages and dialects.[1]

Pre-training ran in three stages. The first used over 30 trillion tokens at a 4K context. The second added 5 trillion tokens weighted toward STEM, coding and reasoning. The third used long-context data to extend the context to 32K.[1]

Post-training had four stages: a long chain-of-thought cold start, reasoning reinforcement learning, thinking-mode fusion, and general reinforcement learning across more than 20 tasks.[1]

Main features

The two MoE models are Qwen3-235B-A22B, with 235 billion total parameters, 22 billion active, 94 layers and 128 experts of which 8 are active, and Qwen3-30B-A3B, with 30 billion total and 3 billion active.[1] The six dense models have 32B, 14B, 8B, 4B, 1.7B and 0.6B parameters.[1]

The context window is 128K tokens for the MoE models and for the 8B to 32B dense models. The 0.6B to 4B models have 32K.[1]

The Qwen team said Qwen3-235B-A22B was competitive with DeepSeek-R1, o1, o3-mini, Grok-3 and Gemini-2.5-Pro on coding, math and general benchmarks.[1] The release also strengthened agent and coding abilities and improved support for the Model Context Protocol (MCP).[1]

Place in development

DeepSeek distilled its R1-0528 model into Qwen3 8B Base, producing DeepSeek-R1-0528-Qwen3-8B.[4]

In July 2025 the Qwen3-2507 releases updated the line and split instruct and thinking models into separate versions. Qwen3 was succeeded by Qwen3.5 in February 2026.[2][3]

CapabilitiesLarge language modelsReasoning

TechniquesFoundation models

Ecosystem & regionOpen sourceChina

Version history

14
  1. 2023
    • Aug 3, 2023 Qwen (Tongyi Qianwen, 2023)

      Qwen, known in Chinese as Tongyi Qianwen (通义千问), is Alibaba's family of large language models. Its first open-weight model, Qwen-7B, was released on 3 August 2023, and the family has grown through many later versions.

  2. 2024
    • Feb 5, 2024 Qwen1.5

      Qwen1.5 is the Qwen family's open model release of February 2024, with base and chat models from 0.5B to 72B parameters, later extended to 110B, and a mixture-of-experts model.

    • Jun 7, 2024 Qwen2

      Qwen2 is the Qwen family's June 2024 generation, with five sizes from 0.5B to 72B parameters, context lengths up to 128K tokens, and most sizes released under Apache 2.0.

    • Sep 19, 2024 Qwen2.5

      Qwen2.5 is the Qwen family's September 2024 generation. Its launch included 13 open-weight models, with up to 18 trillion pretraining tokens, 128K context and most models under Apache 2.0.

    • Nov 28, 2024 QwQ-32B-Preview

      QwQ-32B-Preview is an experimental reasoning research model released by the Qwen team on 28 November 2024, with 32B parameters, a 32K-token context and open Apache 2.0 weights. It scored strongly on maths benchmarks but the team documented clear failure modes.

  3. 2025
    • Jan 28, 2025 Qwen2.5-Max

      Qwen2.5-Max is Alibaba's large mixture-of-experts model, announced in January 2025. It was offered through Qwen Chat and the Alibaba Cloud API, and its weights were not released.

    • Mar 6, 2025 QwQ-32B

      A 32-billion-parameter open-weight reasoning model released by the Qwen team on 6 March 2025, trained mainly through scaled reinforcement learning and reported to perform comparably to the much larger DeepSeek-R1.

    • Apr 29, 2025 Qwen3 · This page

      Qwen3 is a family of large language models from Alibaba, released on 29 April 2025 under Apache 2.0. Its main feature is hybrid thinking, which lets one model reason step by step or answer directly.

    • Jul 21, 2025 Qwen3-2507

      Qwen3-2507 is a staged 2025 refresh of the open-weight Qwen3 models that split thinking and non-thinking into separate checkpoints, raised native context to 262,144 tokens, and scored far above the original non-thinking mode on published benchmarks.

    • Sep 2025 Qwen3-Max

      Qwen3-Max is Alibaba's proprietary trillion-parameter language model, released in September 2025 and served through Alibaba's API. It is the Max-tier flagship that followed Qwen2.5-Max.

    • Sep 2025 Qwen3-Next

      Qwen3-Next is an open-weight model from Alibaba's Qwen team, released in September 2025 under Apache 2.0. It has 80B total parameters, only 3B active per token, and a hybrid attention design later used in Qwen3.5.

  4. 2026
    • Feb 16, 2026 Qwen3.5

      Qwen3.5 is a natively multimodal model family from Alibaba released in February 2026. Its open flagship, Qwen3.5-397B-A17B, has 397B total parameters and 17B active per token, under Apache 2.0.

    • Apr 2, 2026 Qwen3.6

      A generation of Alibaba's Qwen models released in April 2026, spanning proprietary Plus and Max versions aimed at agentic coding and open-weight models such as Qwen3.6-35B-A3B, a 35B-parameter mixture-of-experts with 3B active parameters.

    • Aug 3, 2026 Qwen3.8

      Qwen3.8 is Alibaba's 2026 flagship line. Its open 2.4-trillion-parameter model, Qwen3.8-2.4T-A95B, released in August 2026, has 95B active parameters and a custom license.

20251

  1. ReleaseAlibaba releases Qwen3 with hybrid thinking modes

    On 29 April 2025 Alibaba released Qwen3, a model family with hybrid thinking modes, two mixture-of-experts models and six dense models, under the Apache 2.0 license.

Developed by1

Alibaba Cloud

Foundation for3

Qwen3-VLQwen3-CoderQwen3-Omni

Succeeded by2

Qwen3.5Qwen3-2507

Successor of2

Qwen2.5QwQ-32B

Powers1

Qwen app (Tongyi)

Uses2

Chain of thought and reasoning modelsMixture of experts (MoE)

Notes and sources

  1. Qwen3: Think Deeper, Act Fasterhttps://qwenlm.github.io/blog/qwen3/
  2. QwenLM/Qwen3 repositoryhttps://github.com/QwenLM/Qwen3
  3. Qwenhttps://en.wikipedia.org/wiki/Qwen
  4. deepseek-ai/DeepSeek-R1-0528https://huggingface.co/deepseek-ai/DeepSeek-R1-0528

Cite this page: Qwen3. TellWorld, https://www.tellworld.net/models/qwen3. Updated 2026-11-08