🚀 Qwen3-VL-30B-A3B-Instruct & Thinking are here!
Smaller size, same powerhouse performance 💪—packed with all the capabilities of Qwen3-VL!

🔧 With just 3B active params, it’s rivaling GPT-5-Mini & Claude4-Sonnet — and often beating them across STEM, VQA, OCR, Video, Agent tasks, and more.

And that’s not all: we’re also releasing an FP8 version, plus the FP8 of the massive Qwen3-VL-235B-A22B!

Try it out and make your multimodal AI applications run faster!🧠🖼️

Qwen Chat:   chat.qwen.ai/?models=qwen3-…
Github&Cookbooks：   github.com/QwenLM/Qwen3-V…
API:   alibabacloud.com/help/en/model-…
Blog：  qwen.ai/blog?id=99f033…
ModelScope:   modelscope.cn/collections/Qw…
HuggingFace:   huggingface.co/collections/Qw… Performance of Qwen3-VL-30B-A3B-Thinking

🚀 Introducing Qwen3-Omni — the first natively end-to-end omni-modal AI unifying text, image, audio & video in one model — no modality trade-offs!

🏆 SOTA on 22/36 audio & AV benchmarks
🌍 119L text / 19L speech in / 10L speech out
⚡ 211ms latency | 🎧 30-min audio understanding
🎨 Fully customizable via system prompts
🔗 Built-in tool calling
🎤 Open-source Captioner model (low-hallucination!)

🌟 What’s Open-Sourced?
We’ve open-sourced Qwen3-Omni-30B-A3B-Instruct, Qwen3-Omni-30B-A3B-Thinking, and Qwen3-Omni-30B-A3B-Captioner, to empower developers to explore a variety of applications from instruction-following to creative tasks.

Try it now 👇
💬 Qwen Chat: chat.qwen.ai/?models=qwen3-…
💻 GitHub: github.com/QwenLM/Qwen3-O…
🤗 HF Models: huggingface.co/collections/Qw…
🤖 MS Models:
modelscope.cn/collections/Qw…
🎬 Demo: huggingface.co/spaces/Qwen/Qw… Use the voice chat and video chat features on Qwen Chat to experience the Qwen3-Omni model.

🚀 Introducing Qwen3-Next-80B-A3B — the FUTURE of efficient LLMs is here!

🔹 80B params, but only 3B activated per token → 10x cheaper training, 10x faster inference than Qwen3-32B.(esp. @ 32K+ context!)
🔹Hybrid Architecture: Gated DeltaNet + Gated Attention → best of speed & recall
🔹 Ultra-sparse MoE: 512 experts, 10 routed + 1 shared
🔹 Multi-Token Prediction → turbo-charged speculative decoding
🔹 Beats Qwen3-32B in perf, rivals Qwen3-235B in reasoning & long-context

🧠 Qwen3-Next-80B-A3B-Instruct approaches our 235B flagship.
🧠 Qwen3-Next-80B-A3B-Thinking outperforms Gemini-2.5-Flash-Thinking.

Try it now: chat.qwen.ai
Blog: qwen.ai/blog?id=4074cc…
Huggingface: huggingface.co/collections/Qw…
ModelScope: modelscope.cn/collections/Qw…
Kaggle: kaggle.com/models/qwen-lm…
Alibaba Cloud API: alibabacloud.com/help/en/model-… Pretraining Efficiency & Inference Speed

🚀 Excited to introduce Qwen-Image-Edit!
Built on 20B Qwen-Image, it brings precise bilingual text editing (Chinese & English) while preserving style, and supports both semantic and appearance-level editing.

✨ Key Features
✅ Accurate text editing with bilingual support
✅ High-level semantic editing (e.g. object rotation, IP creation)
✅ Low-level appearance editing (e.g. addition/delete/insert)

Try it now: chat.qwen.ai/?inputFeature=…
Hugging Face: huggingface.co/Qwen/Qwen-Imag…
ModelScope: modelscope.cn/models/Qwen/Qw…
Blog: qwenlm.github.io/blog/qwen-imag…
Github: github.com/QwenLM/Qwen-Im…
API: alibabacloud.com/help/en/model-…

🚀 One line. A full webpage. No hassle.

Introducing Web Dev – the ultimate tool for building stunning frontend webpages & apps using simple prompts in Qwen Chat.

🎨 Just say, "create a twitter website" — and boom! Instant code, ready to go.

No coding required. Just your imagination.

👉🏻 chat.qwen.ai/?inputFeature=… Make an interesting animate

Introducing Qwen3!

We release and open-weight Qwen3, our latest large language models, including 2 MoE models and 6 dense models, ranging from 0.6B to 235B. Our flagship model, Qwen3-235B-A22B, achieves competitive results in benchmark evaluations of coding, math, general capabilities, etc., when compared to other top-tier models such as DeepSeek-R1, o1, o3-mini, Grok-3, and Gemini-2.5-Pro. Additionally, the small MoE model, Qwen3-30B-A3B, outcompetes QwQ-32B with 10 times of activated parameters, and even a tiny model like Qwen3-4B can rival the performance of Qwen2.5-72B-Instruct.

For more information, feel free to try them out in Qwen Chat Web (chat.qwen.ai) and APP and visit our GitHub, HF, ModelScope, etc.

Blog: qwenlm.github.io/blog/qwen3/
GitHub: github.com/QwenLM/Qwen3
Hugging Face: huggingface.co/collections/Qw…
ModelScope: modelscope.cn/collections/Qw…

The post-trained models, such as Qwen3-30B-A3B, along with their pre-trained counterparts (e.g., Qwen3-30B-A3B-Base), are now available on platforms like Hugging Face, ModelScope, and Kaggle. For deployment, we recommend using frameworks like SGLang and vLLM. For local usage, tools such as Ollama, LMStudio, MLX, llama.cpp, and KTransformers are highly recommended. These options ensure that users can easily integrate Qwen3 into their workflows, whether in research, development, or production environments.

Hope you enjoy our new models!
Qwen3 exhibits scalable and smooth performance improvements that are directly correlated with the computational reasoning budget allocated. This design enables users to configure task-specific budgets with greater ease, achieving a more optimal balance between cost efficiency and inference quality.

Announcing Qwen2.5-VL Cookbooks!

🧑‍🍳A collection of notebooks showcasing use cases of Qwen2.5-VL, include local model and API. Examples include Compute use, Spatial Understanding, Document Parsing, Mobile Agent, OCR, Universal Recognition, Video Understanding.

🔗Link:github.com/QwenLM/Qwen2.5…

💬 Qwen Chat: chat.qwenlm.ai (choose Qwen2.5-VL-72B-Instruct as the model)

⚙️ API: alibabacloud.com/help/en/model-… 1 Computer use

This notebook demonstrates how to use Qwen2.5-VL for computer use. It takes a screenshot of a user's desktop and a query, and then uses the model to interpret the user's query on the screenshot.

github.com/QwenLM/Qwen2.5…

The burst of DeepSeek V3 has attracted attention from the whole AI community to large-scale MoE models. Concurrently, we have been building Qwen2.5-Max, a large MoE LLM pretrained on massive data and post-trained with curated SFT and RLHF recipes. It achieves competitive performance against the top-tier models, and outcompetes DeepSeek V3 in benchmarks like Arena Hard, LiveBench, LiveCodeBench, GPQA-Diamond.

📖 Blog: qwenlm.github.io/blog/qwen2.5-m…
💬 Qwen Chat: chat.qwenlm.ai (choose Qwen2.5-Max as the model)
⚙️ API: alibabacloud.com/help/en/model-… （check the code snippet in the blog）
💻 HF Demo: huggingface.co/spaces/Qwen/Qw…

In the future, we not only continue the scaling in pretraining, but also invest in the scaling in RL. We hope that Qwen is able to explore the unknown in the near future! 🔥

💗 Thank you for your support during the past year. See you next year! Results of base language models. We are confident in the quality of our base models and we expect the next version of Qwen will be much better with our improved post-training methods.

🎄Happy holidays and we wish you enjoy this year. Before moving to 2025, Qwen has the last gift for you, which is QVQ!

🎉 This may be the first open-weight model for visual reasoning. It is called QVQ, where V stands for vision. It just reads an image and an instruction, starts thinking, reflects while it should, keeps reasoning, and finally it generates its prediction with confidence! However, it is still experimental and this preview version still suffers from a number of limitations (mentioned in our blog), which you should pay attention to while using the model. Feel free to refer to the following links for more information:

* Blog: qwenlm.github.io/blog/qvq-72b-p…
* HF: huggingface.co/collections/Qw…
* ModelScope: modelscope.cn/models/Qwen/QV…
* Kaggle: kaggle.com/models/qwen-lm…

🚀 It achieves impressive performance in benchmark evaluation, e.g., MMMU, MathVista, etc. But what is more interesting is that it is exciting to see the AI model behaves differently by thinking deeply and reasoning step by step instead of directly providing answers. Yet, it is still a model for preview. It is unstable, it might fall into repetition, it sometimes doesn't follow instruction, etc. We invite you to try the new interesting model and enjoy playing with it! Feel free to shoot us feedback! QVQ achieves significant performance improvements in multiple benchmarks compared with Qwen2-VL-72B-Instruct.

After the release of Qwen2.5, we heard the community’s demand for processing longer contexts.

qwenlm.github.io/blog/qwen2.5-t…

Today, we are proud to introduce the new Qwen2.5-Turbo version, which features:

📚 Longer Context Support: We have extended the model’s context length from 128k to 1M, which is approximately 1 million English words or 1.5 million Chinese characters, equivalent to 10 full-length novels, 150 hours of speech transcripts, or 30,000 lines of code.

🚀 Faster Inference Speed: Using sparse attention mechanisms, we successfully reduced the time to first token for processing a context of 1M tokens from 4.9 minutes to 68 seconds, achieving a 4.3x speedup.

🛢 Lower Cost: The price remains ¥0.3 / 1M tokens. At the same cost, Qwen2.5-Turbo can process 3.6 times the number of tokens as GPT-4o-mini.

Now, you can use it through the API service of
👉🏻 Alibaba Cloud Model Studio [Chinese]: help.aliyun.com/zh/model-studi…,
👉🏻 HuggingFace Demo: huggingface.co/spaces/Qwen/Qw…
👉🏻 ModelScope Demo: modelscope.cn/studios/Qwen/Q… We first conducted experiments on the 1M-token Passkey Retrieval task. The results show that Qwen2.5-Turbo can perfectly capture all hidden numbers in the 1M tokens of irrelevant text, demonstrating the model’s ability to capture detailed information in ultra-long contexts.

🚀Now it is the time, Nov. 11 10:24! The perfect time for our best coder model ever! Qwen2.5-Coder-32B-Instruct!

Wait wait... it's more than a big coder! It is a family of coder models! Besides the 32B coder, we have coders of 0.5B / 1.5B / 3B / 7B / 14B! As usual, we not only share base and instruct models, we also provide quantized models in the format of GPTQ, AWQ, as well as the popular GGUF! 💖

👉🏻Blog: qwenlm.github.io/blog/qwen2.5-c…

👉🏻Tech Report: arxiv.org/abs/2409.12186

👉🏻Hugging Face: huggingface.co/collections/Qw…

👉🏻ModelScope: modelscope.cn/collections/Qw…

👉🏻Kaggle: kaggle.com/models/qwen-lm…

👉🏻GitHub: github.com/QwenLM/Qwen2.5…

👉🏻Demo [chat]: huggingface.co/spaces/Qwen/Qw…

👉🏻 Demo [Artifacts]: huggingface.co/spaces/Qwen/Qw…

The flagship model, Qwen2.5-Coder-32B-Instruct, reaches top-tier performance, highly competitive (or even surpassing) proprietary models like GPT-4o, in a series of benchmark evaluation, including HumanEval, MBPP, LiveCodeBench, BigCodeBench, McEval, Aider, etc. It reaches 92.7 in HumanEval, 90.2 in MBPP, 31.4 in LiveCodeBench, 73.7 in Aider, 85.1 in Spider, and 68.9 in CodeArena! Super excited to launch our models together with one of our best friends, Ollama! Today, the Qwen capybara codes together with Ollama! @ollama

🌟 Meet our mascot, a relaxed capybara. Known for its calmness and adaptability, the capybara symbolizes how Qwen is always ready to help you easily tackle the challenges of the tech world. We hope you like it! Qwen is pre-training

Welcome to the party of Qwen2.5 foundation models! This time, we have the biggest release ever in the history of Qwen. In brief, we have:

Blog:
Blog (LLM):
Blog (Coder):
Blog (Math):
HF Collection:
ModelScope:
HF Demo:

* Qwen2.5: 0.5B, 1.5B, 3B, 7B, 14B, 32B, and 72B
* Qwen2.5-Coder: 1.5B, 7B, and 32B on the way
* Qwen2.5-Math: 1.5B, 7B, and 72B.

All our open-source models, except for the 3B and 72B variants, are licensed under Apache 2.0. You can find the license files in the respective Hugging Face repositories. Furthermore, we have also open-sourced the **Qwen2-VL-72B**, which features performance enhancements compared to last month's release.

As usual, we not only opensource the bf16 checkpoints but we also provide quantized model checkpoints, e.g, GPTQ, AWQ, and GGUF, and thus this time we have a total of over 100 model variants!

Notably, our flagship opensource LLM, Qwen2.5-72B-Instruct, achieves competitive performance against the proprietary models and outcompetes most opensource models in a number of benchmark evaluations!

We heard your voice about your need of the welcomed 14B and 32B models and so we bring them to you. These two models even demonstrate competitive or superior performance against the predecessor Qwen2-72B-Instruct!

SLM we care as well! The compact 3B model has grasped a wide range of knowledge and now is able to achive 68 on MMLU, beating Qwen1.5-14B!

Besides the general language models, we still focus on upgrading our expert models. Still remmeber CodeQwen1.5 and wait for CodeQwen2? This time we have new models called Qwen2.5-Coder with two variants of 1.5B and 7B parameters. Both demonstrate very competitive performance against much larger code LLMs or general LLMs!

Last month we released our first math model Qwen2-Math, and this time we have built Qwen2.5-Math on the base language models of Qwen2.5 and continued our research in reasoning, including CoT, and Tool Integrated Reasoning. What's more, this model now supports both English and Chinese! Qwen2.5-Math is way much better than Qwen2-Math and it might be your best choice of math LLM!

Lastly, if you are satisfied with our Qwen2-VL-72B but find it hard to use, now you got no worries! It is OPENSOURCED!

Prepare to start a journey of innovation with our lineup of models! We hope you enjoy them!qwenlm.github.io/blog/qwen2.5/
qwenlm.github.io/blog/qwen2.5-l…
qwenlm.github.io/blog/qwen2.5-c…
qwenlm.github.io/blog/qwen2.5-m…
huggingface.co/collections/Qw…
modelscope.cn/organization/q…
huggingface.co/spaces/Qwen/Qw… Qwen2.5-72B-Instruct against the opensource models!