Alibaba (Qwen) at a glance
- Model pages in the directory
- 169
- Live on OpenRouter
- 42
- Price range
- $0.03 to $4.00 input, $0.13 to $12.00 output, per million tokens
15 of them served by other hosts
Models priced in the run of 23 Oct 2026
42 priced. Source: OpenRouter models API, 23 Oct 2026
Listed under 5 provider names in the catalogue: alibaba (38), alibaba-nlp (1), qwen (113), tongyi-zhiwen (1), wanx (1). The list below shows qwen.
The provider
Qwen's signature is breadth and openness. It ships the widest ladder of any lab, from tiny on-device models to frontier-class ones, mostly as open weights, and it is strong on multilingual and coding tasks. The house style is prolific and practical. Whatever size you need, there is a Qwen for it.
Where it fits
Teams that want an open-weight model at a specific size or for a specific job, especially multilingual or coding work, with the option to self-host.
Strengths
- The widest ladder of open-weight sizes in the market
- Strong multilingual and coding performance
- Self-hostable for data control
Watch for
- China-based hosted API affects residency for regulated buyers
- Sheer number of variants makes selection harder
- English docs and support lighter than the US labs
Our take
Qwen is the open-weight swiss army knife. Whatever size or job you need, there is likely a strong Qwen for it, and you can self-host. The breadth is the value, though the same breadth means you spend more time choosing the right variant.
The Qwen team at Alibaba Cloud is the most prolific model shop in the market. It ships the widest ladder of any lab, from tiny on-device models to frontier-class ones, and releases most of them as open weights, which has made Qwen a favourite base for builders who want an open model at a very specific size.
The value is breadth. Whatever scale or job you need, from a multilingual chat model to a dedicated coder, there is likely a strong Qwen for it, and you can self-host to control data residency. The same breadth is the one catch: you spend more time picking the right variant than with a lab that ships two or three models.
The catalogue below is every Qwen model we track, joined live to pricing, context, and benchmark scores.
