Model Pricing

All prices are in USD per 1 million tokens. Prices differ by deployment region — International is the default; Global (US/EU) offers lower latency at different rates.

International region

Text models

Model	Input (≤32K)	Input (32–128K)	Input (128–252K)	Output (≤32K)	Output (32–128K)	Output (128–252K)
qwen3-max	$1.20	$2.40	$3.00	$6.00	$12.00	$15.00
qwen-max	$1.60	—	—	$6.40	—	—
qwen3.5-plus	$0.40	—	—	$2.40	—	—
qwen-plus	$0.40	—	—	$1.20	—	—
qwen3.5-flash	$0.10	—	—	$0.40	—	—
qwen-turbo	$0.20	—	—	$0.60	—	—

Models marked with — use flat pricing regardless of context length.

Vision models

Model	Input	Output
qwen-vl-max	$0.80	$3.00
qwen-vl-plus	$0.21	$0.85

Multimodal (Omni)

Model	Modality	Input	Output
qwen3-omni-flash	Text	$0.43	$1.72
qwen3-omni-flash	Audio	$3.00	$6.00

Global region (US / EU)

Lower-latency endpoints. No free tier. Only select models are available.

Model	Input (≤32K)	Output (≤32K)
qwen3-max	$0.359	$1.434
qwen3.5-plus	$0.115	$0.688
qwen3.5-flash	$0.029	$0.287

Batch API pricing

Submit requests via the Batch API to receive 50% off all token prices. Batch jobs complete within 24 hours.

Example: qwen3-max input on International at ≤32K context = $1.20/1M → $0.60/1M with Batch API.

Batch pricing applies to both input and output tokens. See the Batch API guide for how to submit jobs.

Context Caching pricing

When a request reuses a cached prefix (e.g., a long system prompt), the cached portion of input tokens is charged at a reduced rate. The exact discount depends on the model.

Context caching cannot be combined with Batch API pricing — you must choose one or the other per request.

To use context caching, structure your prompt so the stable prefix comes first and mark it with the appropriate cache control parameter. See the Context Caching guide for implementation details.