Tongyi Qianwen

On April 23rd, the Ali Yuntung team announced yesterday that the open source model family had received new members — Qwen 3.6-27B. This is a dense, multi-modular model with 27 billion parameters and the highest community call for model specifications. This was followed by the release of Qwen3.6-Plus and Qwen3.6-35B-A3B, and this open-source version of 27B, while maintaining the advantages of a dense architecture, has resulted in a comprehensive upgrading of intelligent body programming and multimodular reasoning. According to official sources, Qwen 3.6-27B supports..

On December 11th, Ali Thong Yi asked for the release of Qwen3-TTS, an AI model for speech synthesis in support of multi-speaker, multilingual and multi-speaker languages, currently available through Qwen API. 1AI with Qwen3-TTS. The main improvements are as follows: More extensive sound support: Qwen3-TTS provides more than 49 sound colours, covering different genders, age, geographical characteristics and role setting, including ecstasy-trabbits, young pythons-porns, pompous lady-thirty-things, tough teacher-ink lecturers, wise old man..

On 4 December, on the occasion of the International Day of Persons with Disabilities (3 December), the Hangzhou Student Technology Corporation officially launched the country ' s first AI Blinding Glasses. It was described that the smart glasses were based on a series of models such as Qwen-VL, OCR, which had the ability to go for shelter, search for books, voice assistants and one-key help for relatives and friends. In the travel scene, the glasses can achieve a very low delay of 300 ms, which is the equivalent of a road indication for each step. Officially, in order to solve the problem of the travel problems of small visually impaired partners, students will study their own visual models and generalities..

On November 25th, Ariyun, under the umbrella of the Ali Baba Group, announced on November 24 that technical support would be provided for the latest version of the “Sea-Lion” Large Language Model (LLM), known as Qwen-Sea-Lion-v4, of the Singapore National Artificial Intelligence Programme (AI Singapore, ASG). It's worth mentioning that the AISG this time abandoned the Meta model to the Qwen Open Source Architecture of Ali Baba. In a press release, Aliyun stated that the latest version was based on Aliyun's..

On November 14th, according to Wall Street News, Ali Baba has secretly launched the "Major Question" project to launch a full-scale C-end AI application market against the ChatGPT. According to the report, the Ali Baba Group is preparing a major adaptation of its flagship AI application of the "Tun-Tun-Tun-Tun-Tun " to the name "Mun-Qun" App to locate personal AI assistants and synchronize its development of an international version to compete directly for overseas users. According to the information received, Ali Baba has deployed over 100 engineers at Hangzhou headquarters, dedicated to two-storey office buildings to advance research and development. This layout is considered to have announced 380 billion yuan at the beginning of the year..

On 3 November, in the news, Ali Thong Yi asked about the latest flagship model, Qwen3-Max, which is already “deep thinking” on the official web line, a feature that significantly enhances the efficiency of complex task management by enhancing the ability of the chain of reasoning to analyse and dismantle multi-step issues. 1AI noted that Aliyun launched Qwen3-Max on September 24, which is the largest and most capable language model of a communism team to date, with parameters exceeding 1 trillion dollars and pre-training data as high as 36T tokens. Qwen3-Max-Instract Official Edition on behalf of..

On October 13, according to information from 01Founder, two of the country's technology giants, Ali Baba and byte beats, the assistant AI under the banner Qwen and Doubao, began to measure memory functions. This move, widely regarded as the ChatGPT for the leader of the target industry, OpenAI, marks an acceleration of the role of the national production AI assistant from the "Purpose Questions and Answer Tool" to the "Purpose Personal Assistant" role. It is known that OpenAI updated memory for ChatGPT in April this year. Report..

On 25 September, according to the news, at the 2025 Cloud Habitat Congress, the Director of the Ali Group and CEO Wu Yui Song indicated that the 38 billion AI infrastructure was being actively advanced and that additional inputs were planned. At the conference, Ali launched six new models + 1 brands: Qwen MAX: Large Model for trillion Parameters, Coding and Tool Call Capability to Top the International List; Qwen3-Omni: Large Model for a New Generation of Numerical Whole Models, which truly achieves the "Qwen3-VL:Agent and Coding"

On September 24th, following the release of the Qwen3-2507 series, Aliyun announced today the launch of Qwen3-Max, the largest and most capable language model of the general team to date. The Qwen3-Max-Instruct official version further upgrades coding and intelligence capabilities, reaching industry-led standards in comprehensive benchmark tests that cover knowledge, reasoning, programming, command compliance, human preferences, intelligent mission and multilingual understanding. Qwen3-Max-Thi..

September 7 news, Ali quietly in Tongyi Qianqian official website, OpenRouter online Qwen3-Max-Preview model, and called it the most powerful language model in the Tongyi Qianqian series. Afterwards, Tongyi big model through the official microblogging introduced the features that this model has. Compared with the 2.5 series, the released Qwen3-Max-Preview (Instruct) realizes significant enhancements in the dimensions of Chinese and English comprehension, complex instruction following, and tool invocation, and at the same time significantly reduces the knowledge illusion, making the model smarter and more available to...

September 6, 2012 - Ali Tongyi Qianwen has uploaded the latest Qwen-3-Max-Preview model on its official website and OpenRouter. According to the official website, the model is the most powerful language model in the Tongyiqianqian series. 1AI attached the relevant addresses as follows: Official Website: Qwen Chat OpenRouter: Qwen3 Max - API, Providers, Stats The model is described on OpenRouter to the following effect: Input: $1.20 (current exchange rate is about...

August 19th, 2011 - Qwen, a leading provider of image editing solutions, today announced the release of Qwen-Image-Edit, an image editing version of Qwen-Image. Qwen-Image-Edit is further trained based on the 20B Qwen-Image model, and successfully extends the unique text rendering capability of Qwen-Image to the image editing field, realizing the accurate editing of text in images. In addition, Qwen-Image-Edit will input the image into Qwen2.5-...

August 16th news, Tongyi Qianqian official article yesterday, announced that Qwen is upgraded again, the ability to improve again! According to the introduction, Qwen3's 1 million tokens context expansion, Qwen3-30B-A3B-2507, Qwen3-235B-A22B-2507 comprehensive upgrade, with Double Block Attention (DCA) + MInference Sparse Attention, long text is more stable and faster, the performance is up to 3 times higher, and is compatible with vLLM & SGLang and other high-performance deployment frameworks.... SGLang and other high-performance deployment frameworks...

August 9, AliCloud Tongyi Qianqian officially announced that AI programming intelligence body Qwen Code users can get 2000 (request) running privileges for free every day, and there is no Token limit. Its GitHub home page shows that 2000 ModelScope API calls per day are limited to users in mainland China (1000 OpenRouter API calls overseas). In addition, the official GitHub also provides developers with resources for accessing and using GitHub. Qwen Co...

August 5, Ali Tongyi Qianqian team today announced open source Qwen-Image, a 20B MMDiT model, is also the first image generation base model in the Tongyi Qianqian series, which has made significant progress in complex text rendering and accurate image editing. Key features of Qwen-Image include: Text Rendering Capability: Qwen-Image excels in complex text rendering, supporting multi-line layouts, paragraph-level text generation, and fine-grained detail rendering. High fidelity output can be achieved in both English and Chinese. A...

July 31, 2011 - Ali Tongyi Qianqian today announced the launch of a new reasoning model Qwen3-30B-A3B-Thinking-2507. Compared to the Qwen3-30-A3B model open-sourced on April 29th, the new model has significantly improved its reasoning ability, generalization ability, and context length: The new model scored 85.0 points in AIME25, a test that focuses on mathematical ability, and 66.0 points in LiveCodeBench v6, a test of code ability. The new model scored 85.0 in AIME25, a math-focused test, and 66.0 in LiveCodeBench v6, outperforming Ge...

July 23 news, AliCloud Tongyi team today released a notice announcing the official open source of Qwen3-Coder, which is called "intelligent programming engine for the world". This is also the team's most Agent-capable code model to date, in Agentic Coding, Agentic Browser-Use and Foundational Coding Tasks have achieved the SOTA effect of the open-source model, comparable to Claude Sonnet4. According to the presentation...

July 22, 2011 - AliCloud today updated its flagship Qwen3 model with an updated version of Qwen3-235B-A22B-FP8 Non-thinking mode (Non-thinking) named Qwen3-235B-A22B-Instruct-2507-FP8. AliCloud said that after communicating with the community and deep consideration , it decided to stop using the hybrid thinking model and switch to training Instruct and Thinking models separately to get the best quality. According to the introduction, the new Q...

June 28th, Aliyun Tongyi Qianqian just posted a message announcing the launch of the latest Qwen VLo - a multimodal unified understanding and generation model, which users can experience through Qwen Chat (chat.qwen.ai). This newly upgraded model is not only able to "see" the world, but also to create high-quality re-creations based on understanding, truly realizing the leap from perception to generation. According to the introduction, Qwen VLo is able to build the whole picture from left to right, from top to bottom in a gradual and clear way. During the generation process, ...

June 6 news, Ali early this morning open source Qwen3-Embedding series of models (Embedding and Reranker), designed for text representation, retrieval and sorting tasks, based on the Qwen3 base model for training. According to officials, the Qwen3-Embedding family has demonstrated excellent performance in text representation and sorting tasks in multiple benchmark tests. The Qwen3-Embedding family has the following characteristics: Excellent generalization: The Qwen3-Embedding family achieves industry...

Still struggling with which AI tool to choose? DeepSeek, Gemini, Tongyi Thousand Questions, Beanbag, Kimi... Mainstream AI tools on the market are blossoming, with different functions and dazzling publicity and selling points. But there is no "best", only "most suitable". What kind of AI partner do you need for your creative inspiration, office efficiency, code development or knowledge exploration? In this article, we clearly explain the unique advantages and potential shortcomings of the five major tools from the core capabilities, specific scene performance, interactive experience to integration and scalability and other key dimensions. After reading this article, you may no longer be confused and pinpoint...

May 14th, Tongyi Qianqian Qwen official public number yesterday (May 13th) released a blog post announcing the launch of Deep Research, an intelligent assistant system capable of step-by-step planning of complex research tasks, synthesizing a large amount of online information and assisting users in completing complex scientific research tasks. Deep Research combines the vast amount of information available on the Internet with multi-step searches and analysis to summarize complex tasks and present the results in a comprehensive, easy-to-read report. Tasks that used to take hours to complete now take only...

May 1, 2011 - Alibaba continues to push into the AI space, and after releasing the Qwen2.5-Omni-7B model in March, its Qwen team released Qwen2.5-Omni-3B again yesterday (April 30), which is now available for open download on Hugging Face. Note: This 3B parametric model is a lighter version of its 7B flagship multimodal model, designed for consumer-grade hardware, covering a wide range of input functions such as text, audio, image and video. The team says that despite the reduced parametric scale, the 3B...

On April 28th, Ali Tongyi Qianqian open source head Yang Lin Junyan hinted in a post on X that Qwen3 models are expected to be released today. It's worth noting that the Qwen3 model collection was briefly unveiled at ModelScope, Alibaba's open source AI modeling community, and has since been taken offline, including Qwen3-4B-Base, Qwen3-1.7B, Qwen3-0.6B, and Qwen3-30B-A3B-Base. All models are licensed under the Apache License ...