Qwen
Alibaba's open-weights language model family
Overall score
About
Qwen is Alibaba's open-weights large language model family. Includes general-purpose models (Qwen2.5), code models (Qwen-Coder), math models (Qwen-Math), and multimodal variants (Qwen2-VL). Among the strongest open-weights options, particularly outside English-language tasks.
Best for: Builders who need open-weights models with strong multilingual performance (Chinese, Asian languages) or who want a serious alternative to Llama for fine-tuning.
Pricing
Open Weights
- Monthly
- n/a
- Annual /mo
- Free
- Billing
- flat
- Notes
- All Qwen weights;Apache 2.0 licence on most variants;Self-host or use any inference provider · Inference costs are paid to whichever provider hosts the model.
| Tier | Monthly | Annual /mo | Billing | Notes |
|---|---|---|---|---|
| Open Weights | n/a | Free | flat | All Qwen weights;Apache 2.0 licence on most variants;Self-host or use any inference provider · Inference costs are paid to whichever provider hosts the model. |
Key features
- Multiple model sizes (0.5B to 72B+)
- Specialised variants (Coder, Math, VL)
- Strong multilingual benchmarks
- Apache 2.0 licence on most variants
- Available on every major inference provider
Integrations
Trust & compliance
- Stage range
- n/a
- Founded
- 2023
- Status
- active
- SOC 2
- n/a
- GDPR
- n/a
- Data residency
- customer_choice
- External rating
- n/a
- Last verified
- Jul 2026
Reviews
Be the first to share your experience.
Related tools in Agent infrastructure
Pairs well with
People who've discussed Qwen
See all people →Curated mentions from podcasts, posts, and public stacks. Editorial coverage; not endorsements.
Brian Chesky
Co-founder & CEO
Brian Chesky names Alibaba's Qwen model as the workhorse inside Airbnb's customer-service AI stack. The system runs 13 different models in parallel but Chesky has publicly stated Qwen does the bulk of the heavy lifting in production because of its cost and latency profile, with OpenAI's models used more sparingly to balance quality and economics.
Tobi Lütke
CEO & Founder
QMD, Tobi's personal local-search tool, runs three Qwen models under the hood: Qwen3-Embedding-0.6B for embeddings, qwen3-reranker-0.6b for LLM reranking, and a fine-tuned Qwen 1.7B for query expansion. The whole stack runs locally via node-llama-cpp with GGUF model files. It's a useful signal that Tobi reaches for open-weights models when he wants full local control and no cloud dependency.