Qwen
Alibaba Qwen: 1M context + Token Plan from ¥39/mo + OpenAI/Anthropic-compatible APIs
Why we recommend Qwen
Why we recommend Qwen
Qwen (a.k.a. 通义千问 / Tongyi Qianwen) is Alibaba's flagship large-model family.
Since 2023 it has grown into the most complete LLM ecosystem in China:
text, vision, image generation (Wan2.7), video generation (Wan2.7-T2V /
R2V), speech synthesis (TTS), and speech recognition (ASR) — all under a
single GitHub organization (QwenLM, 18.3k followers, 49 repos).
One Alibaba, two domains. Qwen is one LLM family, but the entry points are split:
tongyi.aliyun.com(the "Tongyi Lab" brand portal) is where you read about models, product stories, and partner inquiries — the "Visit official site" button at the top points here;bailian.console.aliyun.com("Bailian" / Model Studio) is the developer console — API keys, Token Plan subscription, and usage billing — the "View plan" buttons on the right jump to the Bailian console. Both run on the same Alibaba Cloud account: sign up once and you can use both.
Qwen3.7 series is the current flagship lineup:
| Model | Context | Positioning |
|---|---|---|
| Qwen3.7-Max | 256K | Complex tasks — strongest capability |
| Qwen3.7-Plus | 1M | Balanced effect / speed / cost (recommended workhorse) |
| Qwen3.7-Flash | 1M | Simple tasks, fast, very cheap |
| Qwen3-Coder-Plus | 1M | Tool calling + environment interaction (code-specialized) |
| Qwen3-VL-Plus | 256K | Vision understanding |
| Qwen3.Omni-Flash | 64K | Native omni-modal (text / image / audio / video) |
Qwen3.8-Max-Preview is already first-wave on the Token Plan, indicating a clearly faster release cadence than Zhipu / DeepSeek.
Token Plan is Alibaba Cloud Bailian (百炼 / Model Studio)'s subscription product for individual developers — the same shape as Zhipu's Coding Plan: fixed monthly fee + Credits quota pool:
| Tier | Monthly (intro) | Credits / 7 days | Agent concurrency |
|---|---|---|---|
| Lite | ¥39 / month | 2,500 | 1-2 |
| Standard (popular) | ¥139 / month | 10,000 | 3-4 |
| Pro | ¥499 / month | 40,000 | 6-8 |
A separate team plan exists (标准 / 高级 / 尊享 tiers) — out of scope for this vendor page.
The killer feature is API portability: Bailian runs OpenAI Chat
Completions, Anthropic Messages, and the DashScope native API
side-by-side. Code on the OpenAI SDK flips to Qwen3.7-Max by changing a
single baseURL; Anthropic-SDK code works too. See the OpenAI-compat
docs
and Anthropic-compat
docs.
In practice, tongyi.aliyun.com is the most reliably reachable entry point from inside China. New signups get 70M–100M free tokens (90-day validity) — enough to try Qwen3.7-Max, Qwen3-Coder-Plus, and Wan2.7-Image end-to-end. The client ecosystem is fully wired: Qwen Code (26.5k⭐ local CLI Agent), Qwen-Agent (16.9k⭐ framework with Function Calling / MCP / RAG), and pre-built Cursor / Claude Code / Cherry Studio / Cline integrations — no extra glue code needed.
If your project needs a full-modality domestic-stack service + 1M-token context + a multi-client ecosystem + predictable subscription budgeting, Qwen's Token Plan is the lowest-friction 2026 pick.