Qwen Research Index

Qwen-Image-2.1: Compact, Efficient, and Unified Image Creation

Sun, 20 Sep 2026 20:00:00 +0800

We are excited to open-source Qwen-Image-2.1, an image model in the Qwen family that balances generation quality, inference efficiency, and cost. Qwen-Image-2.1 unifies text-to-image generation and image editing in a single model, with just 7B parameters in its visual generation component and native support for generating and editing transparent images. This update introduces four key improvements

Qwen3.8-LiveTranslate: Names the speaker. Carries the meaning.

Fri, 18 Sep 2026 17:30:00 +0800

Simultaneous interpretation is not only about translating fast — it must also hear clearly and translate accurately. Qwen3.8-LiveTranslate rebuilds real-time simultaneous interpretation with an Interleave architecture, improving faithfulness, fluency, and conciseness across the board, while average lagging (LAAL) drops from 2.8 seconds to 2.3 seconds. We want simultaneous interpretation to convey

Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery.

Fri, 18 Sep 2026 15:00:00 +0800

Today, we are launching Qwen3.8-Omni-Flash, our next-generation native omnimodal model. Its core objective is to strengthen agent capabilities in real-world productivity scenarios, advancing omnimodal models from “understanding omnimodal content” to “planning tasks, calling tools, and completing creative work.” Building on general agentic capabilities in coding, text-based knowledge work, and GUI

E-Commerce Bench: Long-Horizon Operations, Multi-Dimensional Evaluation

Thu, 03 Sep 2026 10:00:00 +0800

Agent benchmarks over the past few years have mostly followed one pattern. A goal is handed to the model, and the model tries to reach it within a bounded number of turns, whether that means finding the treasure in a maze, producing a report, or fixing a piece of code. Performance is then scored on the quality of the deliverable or on how much of the task got done, and evaluations of this kind usu

Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving

Thu, 03 Sep 2026 08:00:00 +0800

We introduce Qwen-Drive-1.0, the first vision-language foundation model for autonomous driving that unifies 3D perception and visual question answering at the pretraining stage and further extends to motion planning, while keeping the pretrained VLM architecture entirely untouched. Built on the natively multimodal Qwen3.5-4B, it attaches two external modules. A BEV perception head serves as an exp

Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency

Wed, 26 Aug 2026 20:30:00 +0800

In this release we are opening the weights of Qwen3.8-Flash-Next, a multimodal MoE model that also serves as an early preview of the architecture used in Qwen4. It plays the same role that Qwen3-Next played for Qwen3.5: the hybrid Gated DeltaNet + Gated Attention design introduced at that time has since been used across the Qwen3.5, Qwen3.6, Qwen3.7 and Qwen3.8 series. We are again releasing the a

Qwen3.8-Max: A New Bar for Coding and Cowork

Mon, 03 Aug 2026 10:00:00 +0800

Today, we are officially releasing Qwen 3.8-Max, the most capable model in the Qwen family to date. This also marks the first time we will open-source the weights of a Qwen-Max-class model — the open weights will be released next week. Built upon the architectural foundation of Qwen 3.5, Qwen 3.8-Max scales to 2.4 trillion parameters, delivering comprehensive improvements across coding, work, rese

Qwen-Image-3.0: Rich Content, Authentic Details, Deep Knowledge

Tue, 21 Jul 2026 14:00:00 +0800

We are launching Qwen-Image-3.0, the third-generation foundational image generation model in the Qwen-Image series. If the keyword for Qwen-Image-1.0 was "Precision", and the keywords for Qwen-Image-2.0 were "Precision, Variety, Completeness, Beauty, and Authenticity", then the core of Qwen-Image-3.0 comes down to a single word — "Real" (实). This "Real" is embodied across three dimensions: Ric

Qwen-AgentWorld: Language World Models for General Agents

Tue, 23 Jun 2026 11:30:30 +0800

Today we release Qwen-AgentWorld, a native language world model that simulates agent environments across seven domains: Native world modeling: environment modeling is the training objective from continual pre-training onward (CPT → SFT → RL), not a post hoc adaptation on top of a general-purpose LLM. Seven domains, one model: a single model simulates text-based (MCP, Search, Terminal, SWE) and

Qwen-Robot Suite: A Foundation Model Suite for Physical World Intelligence

Tue, 16 Jun 2026 10:00:00 +0800

The Qwen family of foundation models already gives strong perception and reasoning about the physical world. But seeing is not acting: the gap between vision and language understanding and physical control remains the central bottleneck for embodied intelligence. The Qwen-Robot Suite bridges this gap with three foundation models — Qwen-RobotNav, Qwen-RobotManip, and Qwen-RobotWorld. Nav unifies fi

Qwen-RobotNav: A Scalable Navigation Model Designed for an Agentic Navigation System

Tue, 16 Jun 2026 08:00:00 +0800

Agentic navigation systems require a base navigation model with a configurable navigation context protocol: instruction following, object search, target tracking, and autonomous driving share the same perception-planning backbone yet demand fundamentally different context strategies for consuming the visual stream. Like the Model Context Protocol for LLM tool use, a navigation model needs a standa

Qwen-RobotWorld: Boundless Worlds for Embodied Agents

Tue, 16 Jun 2026 08:00:00 +0800

Embodied intelligence requires agents to perceive, reason about, and act within physical environments. World models offer a scalable path forward — but current approaches face a fundamental tension. General video generation models learn rich visual priors but lack the ability to model embodied physics. Domain-specific embodied models are tailored to individual scenarios and cannot generalize acros

Qwen-RobotManip: Alignment Unlocks Scale for Robotic Manipulation Foundation Models

Tue, 16 Jun 2026 08:00:00 +0800

Qwen-Omni × Qwen-RobotManip — Qwen-Omni observes the scene, randomly proposes manipulation tasks via speech, and judges execution in real time. Each video shows Qwen-RobotManip completing tasks on the fly with no pre-defined task list, demonstrating open-ended instruction following and generalization. Qwen-RobotManip is validated across various real-robot platforms and tasks, demonstrating str

Qwen3.7-Plus: Multimodal Agent Intelligence

Mon, 01 Jun 2026 10:00:00 +0800

Today we introduce Qwen3.7-Plus — a multimodal agent model that unifies vision and language into a single, versatile agent foundation. Building on Qwen3.7's strong text backbone, Qwen3.7-Plus delivers a comprehensive upgrade in vision-language capabi

Qwen-VLA: From Understanding the World to Acting in It

Fri, 29 May 2026 17:00:00 +0800

Over the past few years, multimodal large language models have become increasingly capable of understanding images, videos, and real-world scenes. They can recognize objects, reason about spatial relationships, answer visual questions, and solve complex multimodal reasoning tasks. But for embodied intelligence, understanding the world is only the first step. A truly embodied agent also needs to un

Qwen3.7: The Agent Frontier

Wed, 20 May 2026 10:00:00 +0800

Today we introduce Qwen3.7-Max, our latest proprietary model designed for the agent era. Qwen3.7-Max is built to be a versatile agent foundation — equally capable of writing and debugging code, automating office workflows, and sustaining autonomous execution across hundreds or thousands of steps. What sets Qwen3.7-Max apart is the breadth and depth of its agent capabilities. It excels as a codin

Qwen3.5-LiveTranslate: From Sound to Sight, From Word to Right

Tue, 19 May 2026 17:40:00 +0800

Qwen3.5-LiveTranslate-Flash is the latest simultaneous interpretation model in the Qwen family, built on top of Qwen3.5-Omni. It delivers real-time, multimodal translation that not only hears and translates speech, but also sees and understands visual context to produce more accurate translations. Compared with its predecessor Qwen3-LiveTranslate, Qwen3.5-LiveTranslate-Flash brings major upgrades

Qwen-Scope: Decoding Intelligence, Unleashing Potential

Thu, 30 Apr 2026 12:00:00 +0800

Interpretability research has emerged as a critical area for understanding LLM behaviors, informing performance optimization, and enabling more controllable model outputs. Today, we are excited to introduce Qwen-Scope, an interpretability toolkit trained on the Qwen3 and Qwen3.5 series models. Specifically, we inserted and trained Sparse Autoencoders (SAEs) within Qwen’s hidden layers. By imposing

FlashQLA: CP-/Bwd-Friendly Fused Linear Attention Kernels for GDN

Tue, 28 Apr 2026 10:00:00 +0800

Following the release of Qwen3-Next, Gated Delta Network (GDN) has become the workhorse attention layer across the Qwen family — from Qwen3-Next-80B-A3B all the way to the subsequent Qwen3.5 / Qwen3.6 series. As models scale to 397A17B / 122A10B / 35B / 27B and context win

Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model

Wed, 22 Apr 2026 10:00:00 +0800

Following the launch of Qwen3.6-Plus and Qwen3.6-35B-A3B, we are excited to open-source Qwen3.6-27B — a dense 27-billion-parameter multimodal model at the scale the community has been asking for most. Still supporting both multimodal thinking and non-thinking modes, Qwen3.6-27B delivers flagship-level agentic coding performance, surpassing the previous-generation open-source flagship Qwen3.5-397B-

Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving

Sat, 18 Apr 2026 10:00:00 +0800

Following the release of Qwen3.6-Plus, we are sharing an early preview of our next proprietary model: Qwen3.6-Max-Preview. Compared to Qwen3.6-Plus, this preview release brings stronger world knowledge and instruction following, along with significant agentic coding improvements across a wide range of benchmarks. As a preview, the model is still under active development — we are continuing to iter

Qwen3.6-35B-A3B: Agentic Coding Power, Now Open to All

Wed, 15 Apr 2026 10:00:00 +0800

Following the launch of Qwen3.6-Plus, we are excited to open-source Qwen3.6-35B-A3B — a sparse yet remarkably capable mixture-of-experts (MoE) model with 35 billion total parameters and only 3 billion active parameters. Despite its efficiency, Qwen3.6-35B-A3B delivers outstanding agentic coding performance, surpassing its predecessor Qwen3.5-35B-A3B by a wide margin and rivaling much larger dense

Qwen3.6-Plus: Towards Real World Agents

Thu, 02 Apr 2026 04:00:00 +0800

Following the release of the Qwen3.5 series in February, we are thrilled to announce the official launch of Qwen3.6-Plus. Available immediately via our API, this release represents a massive capability upgrade over its predecessor. Most notably, we have drastically enhanced the model's agentic coding capabilities. From frontend web development to complex, repository-level problem solving, Qwen3.6-

Qwen3.5-Omni: Scaling Up, Toward Native Omni-Modal AGI

Mon, 30 Mar 2026 04:00:00 +0800

Qwen3.5-Omni is Qwen’s latest generation of fully omnimodal LLM, supporting the understanding of text, images, audio, and audio-visual content. Both the Thinker and Talker in Qwen3.5-Omni adopt the Hybrid-Attention MoE. Qwen3.5-Omni series includes Instruct versions in three sizes: Plus, Flash, and Light, with support for 256k long-context input. The model can process more than 10 hours of audio i

Qwen3.5-Max-Preview Now Available on Arena

Thu, 19 Mar 2026 04:00:00 +0800

We are pleased to announce the deployment of Qwen3.5-Max-Preview on Arena, where it has demonstrated exceptional performance during the preliminary evaluations. As we proceed with final optimizations ahead of the release within the next two weeks, we invite the community to evaluate the model's capabilities via https://arena.ai/.

Qwen3.5: Towards Native Multimodal Agents

Mon, 16 Feb 2026 04:00:00 +0800

We are delighted to announce the official release of Qwen3.5, introducing the open-weight of the first model in the Qwen3.5 series, namely Qwen3.5-397B-A17B. As a native vision-language model, Qwen3.5-397B-A17B demonstrates outstanding results across a full range of benchmark evaluations, including reasoning, coding, agent capabilities, and multimodal understanding, empowering developers and enter

Qwen-Image-2.0: Professional infographics, exquisite photorealism

Tue, 10 Feb 2026 13:08:30 +0800

We are launching Qwen-Image-2.0, a next-generation foundational image generation model. The key highlights of Qwen-Image-2.0 include: Professional Typography Rendering: Supports 1k-token instructions for direct generation of professional infographics, including PPTs, posters, comics, and more. Stronger Semantic Adherence: Native 2K resolution support for finely detailed realistic scenes, including

Qwen3-Coder-Next: Pushing Small Hybrid Models on Agentic Coding

Tue, 03 Feb 2026 04:00:00 +0800

--- We introduce Qwen3-Coder-Next, an open-weight language model designed specifically for coding agents and local development. Built on top of Qwen3-Next-80B-A3B-Base, which adopts a novel architecture with hybrid attention and MoE, Qwen3-Coder-Next has been agentically trained at scale on large-scale executable task synthesis, environment interaction, and reinforcement learning, obtaining strong

Qwen3-ASR & Qwen3-ForcedAligner is Now Open Sourced: Robust, Streaming and Multilingual!

Thu, 29 Jan 2026 00:00:04 +0800

Qwen3-ASR family includes two powerful all-in-one speech recognition models and a novel non-autoregressive speech forced alignment model. Qwen3-ASR-1.7B and Qwen3-ASR-0.6B are ASR models that support language identifi

Pushing Qwen3-Max-Thinking Beyond its Limits

Mon, 26 Jan 2026 04:00:00 +0800

We present Qwen3-Max-Thinking, our latest flagship reasoning model. By scaling up model parameters and leveraging substantial computational resources for reinforcement learning, Qwen3-Max-Thinking achieves significant performance improvements across multiple dimensions, including factual knowledge, complex reasoning, instruction following, alignment with human preferences, and agent capabilities.

Qwen3-TTS Family is Now Open Sourced: Voice Design, Clone, and Generation!

Thu, 22 Jan 2026 00:00:04 +0800

Qwen3-TTS is a series of powerful speech generation capabilities developed by Qwen, offering comprehensive support for voice clone, voice design, ultra-high-quality human-like speech generation, and natural language-

Qwen3-VL-Embedding and Qwen3-VL-Reranker: For the Next Generation of Multimodal Retrieval

Thu, 08 Jan 2026 04:00:00 +0800

In June 2025, we open-sourced the text-oriented Qwen3-Embedding and Qwen3-ReRanker model series, providing best-in-class performance across a variety of downstream tasks, including multilingual text retrieval, clustering, and classification, which have been widely adopted by developers in the community. Today, we are thrilled to announce the release of the Qwen3-VL-Embedding and Qwen3-VL-Reranker

Qwen-Image-2512: Finer Details, Greater Realism

Wed, 31 Dec 2025 13:08:30 +0800

We are excited to introduce Qwen-Image-2512, the December update of Qwen-Image’s text-to-image foundational model. You are welcome to try the latest model at Qwen Chat. Compared to the base Qwen-Image model released in August, Qwen-Image-2512 features the following key improvements: Enhanced Huamn Realism Qwen-Image-2512 significantly reduces the “AI-generated” look and substantially enhances ov

Qwen-Image-Edit-2511: Improve Consistency

Tue, 23 Dec 2025 13:08:30 +0800

We are excited to introduce Qwen-Image-Edit-2511, an enhanced version over Qwen-Image-Edit-2509, featuring multiple improvements—including notably better consistency. To try out the latest model, please visit Qwen Chat and select the Image Editing feature. Note that the online version includes certain optimizations for speed; for the best possible performance, we recommend deploying the model loc

Qwen3-TTS Steps Up: Voice Cloning and Voice Design!

Tue, 23 Dec 2025 00:00:45 +0800

Qwen3-TTS family has launched two new models: the voice design model Qwen3-TTS-VD-Flash (accessible via the Qwen API) and the voice cloning model Qwen3-TTS-VC-Flash (accessible via the Qwen API). Key Features: Voice Design：Qwen3-TTS-VD-Flash supports complex natural language instructions, enabling fine-grained control over timbre, prosody, emotion, p

Qwen-Image-Layered: Layered Decomposition for Inherent Editablity

Fri, 19 Dec 2025 13:08:30 +0800

Today, we are excited to introduce Qwen-Image-Layered, a model capable of decomposing an image into multiple RGBA layers. This layered representation unlocks inherent editability: each layer can be independently manipulated without affecting other content. Meanwhile, such a layered representation naturally supports high-fidelity elementary operations-such as resizing, reposition, and recoloring. B

Qwen3-Omni-Flash-2025-12-01：Hear You. See You. Follow Smarter!

Tue, 09 Dec 2025 05:00:00 +0800

Qwen3-Omni is a next-generation native multimodal large model capable of seamlessly processing multiple input modalities—including text, images, audio, and video—and generating both text and natural-sounding speech outputs simultaneously via real-time streaming responses. This version introduces mul

SAPO: A Stable and Performant Reinforcement Learning Method for Training Large Language Models

Fri, 05 Dec 2025 04:00:00 +0800

Reinforcement learning (RL) has become a core ingredient in advancing the reasoning capabilities of large language models (LLMs). Modern RL pipelines enable models to solve harder mathematical problems, write complex code, and reason over multimodal inputs. In practice, group‑based policy optimization—where multiple responses are sampled per prompt and their rewards are normalized within the group

Qwen3-TTS Update! 49 Timbres + 10 Languages + 9 Dialects

Fri, 05 Dec 2025 00:00:04 +0800

Qwen3-TTS-Flash is a flagship text-to-speech model that supports multi-timbre, multi-lingual, and multi-dialect speech synthesis. It aims to produce natural and expressive speech and is available via Qwen API. Major Improvements: Richer Timbres Support: Qwen3-TTS offers over 49 high-quality timbres, covering a range of genders, ages, regional traits, and character profiles to meet diverse scenario

Qwen DeepResearch: When Inspiration Becomes Its Own Reason

Thu, 13 Nov 2025 04:59:26 +0800

Click here to experience the latest Qwen DeepResearch _How does inspiration die?_ It usually doesn’t die from “not being good enough”, but from being “too much trouble”. When a thought flashes, it’s still fragile and unverified. After a brief moment of excitement, our brains immediately begin to assess the “cost”: “How ma