State-of-the-art open-source models competitive with the best closed frontier models.
About
Alibaba's Qwen (通义千问) model family has rapidly become one of the most capable open-source series globally. QwQ-32B is a fully open reasoning model competitive with OpenAI o1-mini, while Qwen2.5-72B-Instruct matches Llama 3.1 405B at a fraction of the size. The Qwen series supports 29 languages and extends to vision, audio, and code via purpose-built variants. Models are available via the Alibaba Cloud API and as open weights on Hugging Face.
Key highlights
QwQ-32B is among the strongest open reasoning models, competitive with o1-mini
Qwen2.5-72B beats Llama 3.1 405B on many benchmarks at 5× smaller size
Open-weight models available under Apache 2.0 on Hugging Face
Supports 29 languages with strong bilingual English/Chinese performance
API priced significantly cheaper than comparable Western frontier models
Models
Context windows, pricing, capabilities, and API identifiers — everything you need to choose the right model for your workload.
2025
A fully open reasoning model that applies extended chain-of-thought before answering. Competitive with OpenAI o1-mini on MATH and coding benchmarks — released under Apache 2.0.
qwq-32b2024
Dense 72B model with 128K context and support for 29 languages. Outperforms Llama 3.1 405B on several benchmarks at a fraction of the parameter count. Apache 2.0 licence.
qwen2.5-72b-instruct2025
Alibaba's top commercial model. Delivers frontier-level quality across reasoning, coding, and multilingual tasks at highly competitive pricing.
qwen-max2024
Qwen's flagship vision-language model. Handles multi-image inputs, chart analysis, document OCR, and complex visual reasoning tasks.
qwen-vl-maxUse Cases
Each provider has specialisms shaped by their architecture decisions, training data, and API design. Here is where Alibaba (Qwen) performs best.
Get started
Visit the official Alibaba (Qwen) platform to access API keys, documentation, and pricing details.