Qwen
Alibaba's LLM family: open Qwen3 weights plus closed Qwen3.7-Max
Overview
Qwen is Alibaba's LLM family: the open Qwen3 line ships Apache-2.0 weights free on Hugging Face, while the flagship Qwen3.7-Max is a closed, API-only agent model with a 1-million-token context window. A mid-tier Qwen3.7-Plus adds vision and video input at a fraction of Max's price.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Dense & MoE sizes
- Bilingual CN/EN + code
- Runs on your own GPUs
- Vision + video input
- Tool calling
- Deep reasoning mode
- 1M-token context
- Long-horizon tool use
- List price $2.50/$7.50
Use cases
What you can produce with Qwen
- Write and iterate on original "greenfield" code in 358 programming languages and file formats (Qwen3-Coder) — including complete Android apps or Figma/Framer plugins — without prior coding experience
- Solve complex mathematical problems with step-by-step explanations, with reasoning quality praised by reviewers, suitable for tutoring or exam prep
- Analyse documents up to 1 million tokens (~2,000 A4 pages) and extract structured insights, summaries, or answers from lengthy contracts, research papers, or reports
- Conduct deep research sessions that return citations linked to credible sources, producing balanced research summaries
- Generate natural-sounding multilingual blog posts, creative stories, or marketing copy across 119 supported languages (Qwen3) without an obviously machine-generated tone
- Build and deploy a private on-premise AI assistant for data-sensitive sectors (finance, healthcare, public sector) using open-weights models under the Apache 2.0 licence
- Perform multimodal analysis — describing, interpreting, and answering questions about uploaded images, charts, or screen recordings — to support data analysis or visual content workflows
ASEAN Perspective
Qwen in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Qwen is Alibaba's flagship LLM family and a cornerstone of the open-weight ecosystem — the Qwen3 line ships freely on Hugging Face from small dense models up to large MoE sizes. Capability is frontier-adjacent on coding, reasoning, and multilingual tasks, which keeps it a go-to base for self-hosted and fine-tuned deployments.
The newest flagship, Qwen3.7-Max, is an agent-first model that Alibaba has run autonomously for over 30 hours across 1,000+ tool calls — but it's closed-weight and API-only, priced separately by the token. For ASEAN teams, Qwen is notably strong on Chinese-language work with solid regional cloud presence and competitive API pricing; the trade-off is that the top-tier model now sits behind Alibaba Cloud rather than in the open.
What people say
Qwen's app just became the fastest-growing AI app worldwide — monthly active users jumped 149% to roughly 203 million, putting it in third place globally behind ChatGPT and ByteDance's Doubao. That's the good headline. The uglier one sits on Trustpilot, where Qwen carries a 2.6-out-of-5 "Poor" rating.
The gap between the two numbers tells the actual story. Developers and researchers rate the underlying models well: strong Chinese-English bilingual performance, competitive coding benchmarks, and Apache-2.0 weights that let anyone self-host the open Qwen3 line for free. That crowd is happy.
Consumer-app reviewers are less happy. Recurring complaints: heavy content filtering that blocks a wide range of political topics well beyond China-specific subjects, image generation that lags behind dedicated tools, and responses reviewers call flat or overly hedged next to GPT-class models. One Trustpilot reviewer flatly said Qwen isn't a real alternative to OpenAI's latest.
The split makes sense once you separate the products. The open Qwen3 weights are a genuine developer favorite. Qwen3.7-Max, the new agent-first flagship, is closed and API-only — proprietary infrastructure wrapped in the same free-and-open brand, which is exactly the kind of thing that irritates open-source purists even as it courts enterprise agent workloads. For a team deciding what to build on, that distinction matters more than either headline number.
Summary of public user & expert reviews, compiled by RECATOOLS.
Notable facts
- Qwen2.5-72B achieves performance comparable to Llama 3.1 405B on many benchmarks while being 6x smaller and faster to run.
- Alibaba released Qwen models under the Qwen Community Licence, which permits commercial use without royalties for most companies.
- Qwen2-Audio was the first open-weight model capable of understanding speech, music, and environmental sounds with natural language descriptions.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Qwen's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Qwen unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Qwen directly →
Spotted something out of date? Suggest an update →
More in LLMs & Chat
