Airbnb adopts Alibaba's Qwen AI for customer service chatbot

Airbnb is using Alibaba’s Qwen AI models to support its customer service chatbot, according to remarks by CEO Brian Chesky.

Chesky said Qwen met Airbnb’s requirements better than other options like ChatGPT, citing its speed and cost.

Airbnb, based in San Francisco, uses 13 AI models from multiple providers, including OpenAI and Google, to develop its service bot.

Qwen, developed by Alibaba, is an open source AI system.

Chinese tech firms, including Alibaba and DeepSeek, have recently released several open source AI models.

Chesky’s comments come as open source AI models gain traction among global companies.

🧠 Food for thought

Implications, context, and why it matters.

Many Qwen variants on Hugging Face (a popular hub for hosting and sharing AI models) use Apache 2.0 (a permissive open‑source license that permits commercial use when running models locally), though some families use non‑commercial or custom terms 1 2 3.
Qwen Chat is Alibaba’s hosted chat interface for Qwen. Its Terms of Service restrict commercial use of outputs from the hosted service without prior written consent, even though users get rights to their outputs 4.
The result is a split between self‑hosting (running models in their own environment) and the hosted chat. Companies can use Apache 2.0 variants commercially when they run models locally, but Alibaba‑hosted Qwen Chat may carry non‑commercial limits unless they get permission 1 2 3 4.

AWS Bedrock, AWS’s managed service for accessing foundation models, now hosts Qwen3 with security controls and automatic scaling so teams skip running their own stack 5 6.
Predibase (a platform for private deployment and fine‑tuning of large language models) enables private cloud runs on AWS, Google Cloud Platform (GCP), or Microsoft Azure with data privacy controls 7.
Qwen3‑8B (roughly 8 billion parameters) runs on one Nvidia H100 data‑center GPU. Qwen3‑235B (roughly 235 billion parameters) needs four H100s, which pushes costs up 7.
Vendors that offer fine‑tuning services (adapting a base model with domain‑specific data) help teams tailor Qwen for uses like legal or medical work 7.