📍 Home › Tools

Qwen Studio Review: The Free All-Rounder Nobody Puts a Paywall On

Tools  Updated 2026-11-04 · 3 min read · #chatbot #multimodal #free #open-weights

Status board (checked 2026-11-04)

What it is

Alibaba’s consumer assistant for the Qwen model family. You sign in with email, Google, GitHub or Apple (no phone number, no card — and you can even try it logged out). The 2026 lineup in the model picker: Qwen3.8-Max, the flagship released Aug 3, 2026 — a ~2.4-trillion-param MoE with 1M-token context and, for the first time in the Max line, open weights — plus the fast Qwen3.8-Flash (Aug 26), the Qwen-Image 3.0 drawing model (Aug 4) and Wan 3.0 video (Aug 24).

Free tier

• Global web + apps: free, with no consumer paid tier at all — limits exist but their numbers aren’t published.
• The catch isn’t the wallet, it’s the data: a China-headquartered company runs the servers. And the domestic 千问 app quietly introduced its first paid tiers on 2026-08-10 (office-agent and video generation) — the global web app hasn’t shown equivalent gates so far.

Getting started (3 steps)

1. Open chat.qwen.ai, sign in with Google/GitHub/Apple/email.
2. Pick a model above the input box: Max for complex tasks and whole-book contexts, Flash for quick everyday ones.
3. Flip the switches under the box: thinking mode, web search, Deep Research, image/video creation, Web Dev (describe a page, get a working one). The paperclip takes documents — up to 5 files, 20 MB each per the app’s own config.

Best use

Honest weaknesses

• Daily limits exist but Alibaba doesn’t publish the numbers — you find them by hitting them.
• Confident tone on topics with political gravity in CN context; verify contested claims.
• English is good, not natively-first — some flourishes and examples read translated.

FAQ

Is it really $0 with no subscription?

Yes, as of this check: the global app shows no consumer tier (its App Store page lists no in-app purchases). API is separate — developers pay per token (flagship around $2/$6 per million, OpenAI-compatible).

Can I run the same model myself?

Partly, and that’s the party trick: the flagship’s open-weights release plus the 27B Apache-2.0 sibling and the 125B Flash-Next family all run at home — see our Local AI section.

Is my data going to China?

Requests go to Alibaba’s servers. Same rule as every tool here: read the policy before pasting anything sensitive.