Is Qwen Available in the US? (2026)

Yes, Qwen is available in the US: Alibaba Cloud Model Studio International serves it from a Singapore endpoint with roughly 1M free tokens for 90 days and a US (Virginia) Global deployment, and OpenRouter routes the same models without an Alibaba account. The harder question for US buyers is not access but compliance, because Qwen is built by a China-parent company and where your data lands depends on which endpoint you pick.

Qwen is a family of open and hosted large language models from Alibaba Cloud, and the access story has two clean paths. You can go direct through Alibaba Cloud Model Studio International, or you can reach the same model names through OpenRouter, which proxies the request and bills you in USD without you ever signing a Model Studio contract. Each path carries a different data-residency and billing profile, and that difference is what most "is it available" articles skip.

Is Qwen available in the US (verified 2026-06)

According to Alibaba Cloud Model Studio pricing, Qwen is reachable to US developers through Alibaba Cloud Model Studio International, which runs a Singapore endpoint carrying a free quota of about 1 million tokens for 90 days and a separate US (Virginia) Global deployment that ships with no free quota. The practical read is simple: Singapore is where you prototype cheaply, and Virginia is where you serve US users with lower latency once you are paying anyway. We mark these official figures needs_native_reverify because they are sourced from Alibaba documentation and not yet re-checked on a native DashScope key.

Access path	Region	Free quota	Billing	Best for
Model Studio International	Singapore	~1M tokens / 90 days	Alibaba account (USD)	Cheap prototyping, free trial
Model Studio International	US (Virginia) Global	None	Alibaba account (USD)	US-user latency, production
OpenRouter routing	Provider-dependent	None	OpenRouter (USD)	No Alibaba contract, fast onboarding

There is no US ban on calling Qwen. It is a commercial Alibaba Cloud product with international endpoints, not a geofenced service. What you are choosing between is who holds the contract and where the inference runs.

What it actually costs to reach Qwen from the US

The pricing surprise is real and it favors US buyers. According to Alibaba Cloud Model Studio pricing, Qwen3-Max on the International endpoint is tiered by input size and runs around $2.40 per million input tokens and $12.00 per million output for the 32K to 128K bracket. Routed through OpenRouter, the published rate for the same flagship is about $0.78 input and $3.90 output. That is the inversion buyers miss: for the flagship model, the third-party route can cost materially less than Alibaba's own International endpoint. Treat the Alibaba figure as pending native re-verification, but the gap is wide enough that it is worth pricing both paths before you commit.

The cheaper Qwen tiers stay cheap on either path. The OpenRouter-routed rate for Qwen-Plus sits near $0.26 input and $0.78 output per million, and Qwen3-Coder near $0.22 input and $1.80 output. If you only need a workhorse model rather than the flagship, the absolute cost is low enough that the residency question, not the price, should drive your endpoint choice.

We measured Qwen from the US (first-hand evidence)

Published rate cards tell you the list price. We wanted the billed cost on a real call, so we exercised the Qwen tiers through OpenRouter on 2026-06-26 and logged the token counts and latency the provider actually returned. Every number below is OpenRouter-routed, not an Alibaba native measurement, and that disclosure matters for anyone comparing endpoints.

On the qwen3-coder tier, a coding prompt of 36 input tokens producing 58 output tokens billed $0.0000656 and returned in 1.50 seconds, served as qwen3-coder-480b-a35b. That was the fastest and cheapest call in our set. The flagship qwen3-max, on a 38-in/63-out general prompt, billed $0.00027534 and took 3.22 seconds, which is roughly four times the cost of the coder tier for a similar-length answer.

Tier (OpenRouter-routed)	Tokens (in/out)	Billed cost	Latency
qwen3-coder	36 / 58	$0.0000656	1.50s
qwen3-max (general)	38 / 63	$0.00027534	3.22s
qwen3.5-flash	40 / 2831	$0.00073866	17.85s

The genuine caution from the run was qwen3.5-flash. On a two-sentence prompt it generated 2,831 output tokens, which made it both the slowest call at 17.85 seconds and the most expensive at $0.00073866. The "flash" label did not mean fast or cheap here; verbose output dominated the bill. We re-read the usage object to be sure, and the token count held. If you route to a flash tier expecting a snappy short answer, cap your output tokens or you will pay for an essay.

How Qwen compares to DeepSeek for US buyers

US teams evaluating a China-built model usually weigh Qwen against DeepSeek, and the two answer different needs. According to DeepSeek API Docs, DeepSeek V4-Flash bills $0.14 per million input tokens and $0.28 per million output on its own official endpoint, which is a first-party rate rather than a routed one. Our DeepSeek anchor call confirmed the model resolves and answers a 27-in/120-out coding prompt in 3.22 seconds at that official rate.

The asymmetry to keep straight: our DeepSeek number came from the official api.deepseek.com endpoint, while every Qwen number above is OpenRouter-routed because Qwen was not reachable on a native key in our environment. So the comparison is honest on price magnitude but not endpoint-symmetric. For raw token economics on a standard chat or coding workload, DeepSeek's first-party $0.14/$0.28 undercuts the routed Qwen-Plus and most Qwen tiers; for breadth of model sizes and the open-weight option, Qwen's family is wider. Choose DeepSeek if you want the cheapest first-party rate and a single endpoint. Choose Qwen if you need a specific model size, the Singapore free quota to trial, or an open-weight checkpoint you can self-host.

For the full Qwen rate card across every tier, see the Qwen API pricing hub. For the DeepSeek side of this comparison, see the DeepSeek API pricing hub.

The compliance question regulated buyers must answer

Data residency is the real gate for regulated US buyers, not access. The US (Virginia) Global deployment keeps inference in a US region, which is the answer auditors usually want to hear, while the Singapore endpoint processes data offshore. Either way the model is operated by an Alibaba Cloud entity with a China parent, so a buyer in healthcare, finance, or government should route the residency and vendor-nationality questions through their own compliance review before production, exactly as they would for any DeepSeek deployment.

FAQ

Is Qwen available in the United States? Yes. Alibaba Cloud Model Studio International serves Qwen from a Singapore endpoint with about 1 million free tokens for 90 days and from a US (Virginia) Global deployment with no free quota, and OpenRouter routes the same models in USD without an Alibaba account.

Is there a free way to try Qwen from the US? Yes. According to Alibaba Cloud Model Studio pricing, the Singapore International endpoint carries a free quota of roughly 1 million tokens for 90 days. The US Virginia Global deployment has no free quota, so use Singapore for the trial and Virginia for US-latency production.

Is Qwen cheaper than DeepSeek? For standard workloads, DeepSeek V4-Flash's first-party rate of $0.14 input and $0.28 output per million undercuts most Qwen tiers. Qwen's advantage is breadth of model sizes and an open-weight option, not lowest token price.

Does using Qwen from the US raise compliance concerns? The model is operated by an Alibaba Cloud entity with a China parent. The US Virginia Global deployment keeps inference in a US region, but regulated buyers should still run their own data-residency and vendor-nationality review before production.

Are the Qwen prices here official Alibaba rates? No. Our first-hand cost and latency figures are OpenRouter-routed and disclosed as such. The official Alibaba Model Studio rates we cite are sourced from documentation and marked pending native DashScope re-verification.

This is part of the Qwen API pricing hub, where the full tier-by-tier rate card lives.

Author: Kevin Fan, Customer Success Manager at China LLM Directory, specializing in Chinese LLM ecosystem pricing and cross-border AI compliance. Last verified: 2026-06-26.