MODELS
Qwen 3
Open-weight multilingual model with vision and tool use at budget pricing.
Specs
- Context window
- 128,000
- Max output
- 16,384
- Modalities
- text, image
- Tool use
- ✓
- Vision
- ✓
- Streaming
- ✓
- License
- apache-2.0
- Released
- 2025-04-01
Pricing
- Input / 1M
- $0.40
- Output / 1M
- $1.20
Cost estimate
Alibaba's Qwen 3 is an Apache-2.0 licensed model handling text and images with a 128K context window and 16K max output. It supports tool use and streaming, and is priced at $0.4/M input and $1.2/M output — well below most closed peers. Strong Chinese and English coverage make it a default pick for builders deploying across Asian markets, and the open weights allow self-hosting.
Editor's verdict
Pick Qwen 3 when you need open weights with vision and a Chinese-friendly tokenizer, or when API budget is tight. It won't match GPT-5 or Claude on hard reasoning and long-form coding, and the 16K output cap is restrictive for agent traces or long documents. But for multilingual chat, RAG, and routine tool-calling workloads — especially deployed on your own GPUs — the price-to-capability ratio is hard to beat.
Reviews
No reviews yet. Be the first.
Last updated: 2026-04-29