Provider pricing hub
Qwen API Pricing (July 2026)
The Qwen API costs $0.60/$3.60 per million input/output tokens for the flagship Qwen3.5 397B, $0.40/$2.40 for Qwen3.5 Plus (1M context), and $0.10/$0.40 for Qwen3.5 Flash on Alibaba Cloud Model Studio. Most Qwen models are also open weight, so self-hosting is the zero-per-token alternative.
Alibaba prices Qwen aggressively: the flagship 397B mixture-of-experts model costs less per token than most Western budget tiers, and the Plus and Flash tiers carry 1M-token context windows at prices that undercut almost everything except DeepSeek. The table below lists every Alibaba model in BenchLM's registry, synced from Alibaba Cloud Model Studio's official pricing page (International deployment mode).
For how Qwen performs for the money, see the Qwen model rankings and the Chinese model leaderboard, where Qwen3.5 397B (Reasoning) is among the strongest rows.
Every Qwen API price per 1M tokens
| Model | Input $/M | Cached input $/M | Output $/M | Context |
|---|---|---|---|---|
| Qwen3.5 397B | $0.6 | $0.06 | $3.6 | 128K |
| Qwen3.5 Plus | $0.4 | $0.04 | $2.4 | 1M |
| Qwen3.5 Flash | $0.1 | $0.01 | $0.4 | 1M |
Prices are Alibaba Cloud Model Studio International-mode rates for standard prompt lengths; Qwen3.5 Plus rises to $0.50/$3.00 above 256K input tokens per Alibaba's pricing page. Alibaba does not publish a flat batch-discount tier, so batch columns are omitted. $0/$0 self-host-only rows are excluded.
Estimate your monthly Qwen API bill
| Model | Est. monthly cost |
|---|---|
| Qwen3.5 397B | $13.2 |
| Qwen3.5 Plus | $8.8 |
| Qwen3.5 Flash | $1.8 |
Estimates use Qwen's standard per-token rates from the table above. For task-level presets and cross-provider comparison, use the full AI cost calculator.
How Alibaba tiers Qwen pricing
Three hosted tiers cover the lineup. Qwen3.5 397B ($0.60/$3.60) is the flagship — its reasoning variant posts the strongest Qwen scores on BenchLM's leaderboard. Qwen3.5 Plus ($0.40/$2.40) trades a little capability for a 1M-token context window, with a documented step-up to $0.50/$3.00 above 256K input tokens. Qwen3.5 Flash ($0.10/$0.40) is the high-volume tier for classification, extraction, and summarization.
The pricing story has a second half: much of the Qwen family ships open weight. If your volume is high and steady, downloading the weights and running them on rented GPUs can beat even these per-token rates — see the open-source rankings and local LLM guide.
Qwen vs DeepSeek and Kimi pricing
Within the Chinese-lab tier the fight is close. DeepSeek V4 Flash ($0.14/$0.28) undercuts Qwen3.5 Flash on output; Qwen3.5 Plus ($0.40/$2.40) sits below Kimi K2.5 ($0.60/$3) and well below DeepSeek V4 Pro ($1.74/$3.48) on input. All three providers are an order of magnitude cheaper than Western flagships — the decision usually comes down to benchmark fit per task, not price.
Compare the full tables on the DeepSeek and Kimi hubs, or see how the models themselves rank on the Chinese model leaderboard.
Which Qwen should you use?
Start with Qwen3.5 Flash for anything high-volume — at $0.10/$0.40 with a 1M context window it is one of the cheapest hosted models with a serious context budget. Move to Qwen3.5 Plus when quality gaps show up in production, and reserve Qwen3.5 397B (or its reasoning variant) for coding, math, and agentic work where it competes with models several times its price. For self-hosting, the smaller open-weight Qwen rows in the local LLM rankings run on single-GPU hardware.
Qwen API pricing FAQ
How much does the Qwen API cost?
On Alibaba Cloud Model Studio (International mode), Qwen3.5 397B costs $0.60/$3.60 per million input/output tokens, Qwen3.5 Plus $0.40/$2.40 (rising to $0.50/$3.00 above 256K input), and Qwen3.5 Flash $0.10/$0.40. Older Qwen rows and open-weight variants appear in the full table above.
Is Qwen free to use?
The hosted API is pay-per-token, but most Qwen models are open weight — you can download and run them on your own hardware at no per-token cost. For which variants are worth self-hosting, see the open-source rankings and local LLM guide.
What is Alibaba's Qwen API pricing per million tokens?
Alibaba bills per million tokens with separate input and output rates per model, in deployment-mode tiers (International rates shown here). The full per-token table above is synced from Alibaba Cloud Model Studio's official pricing page on every site build.
How does Qwen compare to DeepSeek on price?
They bracket each other. DeepSeek V4 Flash ($0.14/$0.28) beats Qwen3.5 Flash ($0.10/$0.40) on output but not input; Qwen3.5 397B ($0.60/$3.60) is cheaper than DeepSeek V4 Pro ($1.74/$3.48) on input and pricier per output token. Benchmark fit for your task should decide it — see the Chinese model leaderboard.