Qwen API pricing: Qwen3.8 Max, Plus, Flash, 27B and the open models at every provider

Updated Oct 29, 2026

Qwen is Alibaba's family of models, and it is the most widely resold one in the catalog: more than 600 Qwen offers across dozens of providers. Most Qwen models have open weights, so many providers run them themselves and set their own price - the same model can cost several times more at one provider than at another. Below: the typical price (the median across providers) of the models most of them sell, and live tables of every seller. The tables are rebuilt from today's prices on every page view.

The models in one list

Price per 1M tokens, input / output, in the 28 October 2026 collection; "typical" is the median across providers:

• Qwen3.8 Max - typically $2 / $6, input from $1.30; 1M tokens of context, 27 providers. The flagship.
• Qwen3.8 2.4T-A95B - typically $2 / $6; the open-weight giant, 24 providers.
• Qwen3.7 Max - typically $1.71 / $5.14, 20 providers.
• Qwen3.8 27B - typically $0.40 / $2.50; 1.1M context, 45 providers - the most widely sold Qwen model.
• Qwen3.7 Plus - typically $0.33 / $1.35, input from $0.23; 19 providers.
• Qwen3.8 Flash - typically $0.15 / $0.47, input from $0.10; 1M context, 24 providers. The cheap one for volume.
• Qwen3 Coder 480B - typically $0.39 / $1.60, input from $0.22; 20 providers.
• Qwen3.5 397B-A17B - typically $0.58 / $3.60, 24 providers.
• Qwen3.6 35B-A3B - typically $0.17 / $1, 24 providers; a small mixture-of-experts model, fast and cheap.

Qwen3.8 Max: every provider

Live data: this table is built from today's prices.

Qwen3.8 27B: every provider

Live data: this table is built from today's prices.

Qwen3.8 Flash: every provider

Live data: this table is built from today's prices.

Why the same Qwen costs so differently

• Open weights: any provider can host Qwen3.8 27B, 35B or 397B, so prices follow each provider's own hardware and margin - Qwen3.8 27B is sold by 45 providers at very different prices.
• Quantization: some providers run a compressed version (FP8, INT4) and sell it cheaper; the card says so when the provider publishes it.
• Context: a few providers cut the context to save memory - check the context column before choosing the cheapest row.

How to choose

• The strongest Qwen: Qwen3.8 Max - compare the providers, input goes from $1.30 to $2 and more.
• A general model at a low price: Qwen3.8 27B or Qwen3.7 Plus.
• Volume - classification, extraction, chat at scale: Qwen3.8 Flash.
• Code: Qwen3 Coder 480B, or the other coding models.
• All Qwen offers in one list: Qwen in the catalog.

All prices are in USD per 1 million tokens, without taxes, as the providers publish them; each row links to its card with the source and the date it was checked.