FAMILY / ALIBABA CLOUD

Qwen

An open multilingual family spanning dense, MoE, and hybrid-reasoning models.

First model: 2023
Models: 25
Latest dated version: QwenCloud Decision Model Preview

01 / RELEASE SEQUENCE

Family release sequence

Compare versions →

02 / GENERATIONS

What changed

Qwen-7B

Qwen-7B established a continuously expanding Chinese open-model family for developers.

2023-08-03

Qwen1.5-72B

The 72B Qwen1.5 tier with improved multilingual capability and a 32K context window.

2024-02-04

Qwen2-72B

Qwen2 expanded multilingual and coding capabilities; the 72B tier uses the Tongyi Qianwen license, unlike the smaller Apache 2.0 models.

2024-06-07

Qwen2.5-72B

The largest general-purpose dense Qwen2.5 tier, improving coding, mathematics, and structured output; the 72B tier is not Apache-licensed.

2024-09-19

Qwen2.5-Coder-32B

A representative coding branch of Qwen2.5, with 32.5B parameters and Base and Instruct weights.

2024-11-12

Qwen2.5-VL

A vision-language release group with 3B, 7B, and 72B tiers for image, document, and video understanding.

2025-01-26

QwQ-32B

A 32B open reasoning branch using reinforcement learning to improve complex problem solving.

2025-03-06

Qwen2.5-Omni-7B

Uses a Thinker-Talker architecture to understand text, images, audio, and video and stream text and speech.

2025-03-27

Qwen3-235B-A22B

Qwen3 brought hybrid reasoning and efficient MoE together in an Apache 2.0 open-weight model.

2025-04-29

Qwen3-Coder-480B-A35B

An agentic coding MoE with 480B total and 35B active parameters; native 256K context can be extrapolated to 1M.

2025-07-22

Qwen3-Next-80B-A3B

An 80B-total, 3B-active architecture preview combining hybrid linear attention and ultra-sparse MoE.

2025-09-11

Qwen3-Max

A trillion-plus-parameter Qwen3 flagship for coding, agents, and million-token professional work.

2025-09-24

Qwen3.5-397B-A17B

The first Qwen3.5 flagship, extending the Qwen3-Next hybrid architecture into a native multimodal agent model.

2026-02-16

Qwen3.6-35B-A3B

The efficient Qwen3.6 MoE tier, using 3B active parameters for multimodal, coding, and tool tasks.

2026-04-16

Qwen3.6-27B

The 27B dense Qwen3.6 tier for local deployment and general multimodal agents.

2026-04-22

Qwen3.8-2.4T-A95B

Qwen3.8’s open checkpoint with 2.4T total and 95B active parameters, native 262,144 context, extensible to about 1M.

2026-08-02

Qwen3.8-Max API

Service capabilities and open-checkpoint parameters, weights, and licensing are recorded separately so product limits are not mistaken for checkpoint specifications.

2026-08-02

Qwen3.8-27B

The 27B dense open Qwen3.8 model for local deployment, vision, coding, and tool use.

2026-08-14

Qwen3.8-Flash-Next

It introduced QSA, Gated Residual, and N-gram memory to reduce compute while advancing long-context agents.

2026-08-26

Qwen3.8-Max-0902

It turns the flagship release into a traceable service snapshot, concentrating the update on coding, cowork, agent orchestration, and multimodal understanding.

2026-09-02

Qwen3.8-LiveTranslate Flash Realtime

It extends Qwen3.8 from general omni-modal and agent workflows into a dedicated low-latency speech-translation branch.

2026-09-17

Qwen3.8-Omni-Flash

It extends Qwen3.8-Flash-Next agentic capability into unified audio-video understanding and multimodal workflows with DashScope and OpenAI protocol access.

2026-09-18

Qwen-Audio 3.1 Realtime Plus

It advances Qwen’s audio line into a fuller voice-agent interface with additional system voices and a longer context.

2026-09-20

Qwen3.8-Omni-Flash-Realtime

It extends Qwen3.8 omni-modal capability from file understanding into real-time audio/video dialogue and MCP tool workflows.

2026-09-21

QwenCloud Decision Model Preview

A structured decision model for high-frequency business decisions, supporting classification, yes/no decisions, and scoring with probabilities and confidence; it returns structured decisions rather than natural-language text.

2026-09-24