Qwen
Alibaba Cloud's LLM family, open weights through to the Qwen3.8-Max flagship
Overview
Qwen is Alibaba Cloud's family of large language and multimodal models, offered as open weights, as hosted APIs through Model Studio and Qwen Cloud, and as the free Qwen Chat app. The flagship Qwen3.8-Max, a 2.4-trillion-parameter mixture-of-experts model, had its weights published in August 2026 under a custom licence, while smaller models such as Qwen3.8-27B are Apache 2.0.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete.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.
- Qwen3.8-27B, Qwen3.6 and Qwen3.5 under Apache 2.0
- Qwen3.8-2.4T-A95B under the Qwen3.8-Max licence
- Qwen3.8-Flash-Next under the Qwen Community License 1.0
- 1M-token context
- Batch inference discount
- Context caching
- 1M-token context
- Context caching
- Function calling and web search
- Thinking-mode output $4 per M
- 1M-token context
- 1M-token context
- Thinking and non-thinking modes
- Cached input $0.25 per M
- Built-in web search and code interpreter
API pricing (pay as you go)
| Model | Input | Output |
|---|---|---|
| qwen3.8-max Single tier 0<Token≤1M. | $2 | $6 |
| qwen3.7-max Single tier 0<Token≤1M. | $2.50 | $7.50 |
| qwen3.6-max-preview Lowest tier 0<Token≤128K; 128K-256K is $2 / $12. | $1.30 | $7.80 |
| qwen3-max Lowest tier 0<Token≤32K; up to $3 / $15 at 128K-256K. | $1.20 | $6 |
| qwen3.7-plus List price, lowest tier 0<Token≤256K; the page shows a limited-time 20% discount on top. 256K-1M is $1.2 / $4.8. | $0.40 | $1.60 |
| qwen3.6-plus Lowest tier 0<Token≤256K; 256K-1M is $2 / $6. | $0.50 | $3 |
| qwen3.8-flash Single tier 0<Token≤1M. | $0.15 | $0.47 |
| qwen3.7-flash Lowest tier 0<Token≤32K; up to $0.20 / $0.80 at 256K-1M. | $0.03 | $0.13 |
| qwen-turbo Non-thinking output; thinking-mode output is $0.5. The vendor says Turbo is no longer updated. | $0.05 | $0.20 |
| qwen3-coder-plus Lowest tier 0<Token≤32K; up to $6 / $60 at 256K-1M. | $1 | $5 |
| qwen3-coder-flash Lowest tier 0<Token≤32K; up to $1.6 / $9.6 at 256K-1M. | $0.30 | $1.50 |
Use cases
What you can produce with Qwen
- A self-hosted chat or coding assistant on Apache 2.0 weights such as Qwen3.8-27B, with no per-token fees
- Analysis of documents, codebases or long videos up to a million tokens through the hosted Max and Flash models
- Answers grounded in live web results through the built-in web search tool
- Coding-agent sessions in Qwen Code, Claude Code or Cline through Qwen Cloud's OpenAI- and Anthropic-compatible APIs
- Multilingual output across the 201 languages and dialects the Qwen3.5 generation supports
- Fine-tuned domain models trained on your own data from an open Qwen base
Qwen is the most practical open-weight family to build on in 2026: Apache 2.0 models from under a billion to hundreds of billions of parameters, and now even the Max-tier flagship available to download. Hosted pricing in Alibaba Cloud's Singapore region is low, from US$0.03 per million input tokens, with a 90-day free quota.
The catch is licensing drift. Qwen3.8-Max and Qwen3.8-Flash-Next carry custom licences that restrict companies selling model access or AI assistants, so check the licence of the exact model before you ship.
What we found
Qwen is the large language model family built by Alibaba Cloud, first released to the public in 2023 and now one of the most widely used open-weight model lines. It comes in three forms that are easy to confuse: open weights on Hugging Face and ModelScope, paid hosted models through Alibaba Cloud Model Studio and Qwen Cloud, and a consumer chat app. QuestMobile counted 167 million monthly users for the Qwen app in China by May 2026, second to ByteDance's Doubao.
The biggest change this year is that the flagship is now open. Alibaba unveiled Qwen3.8-Max on 2 August 2026 and, within a week, published the weights of the 2.4-trillion-parameter mixture-of-experts model (95 billion active) as Qwen3.8-2.4T-A95B. DeepLearning.AI's The Batch noted that the downloadable version handles text only and does not carry the hosted model's full million-token context. Earlier versions of this page called it closed and API-only, which was wrong.
Licences now differ by model, so read them. Qwen3.8-27B, Qwen3.6 and Qwen3.5 are Apache 2.0. The Qwen3.8-Max licence requires a separate agreement once a company selling model access or an AI coding or office assistant passes US$50 million in revenue over twelve months. Qwen3.8-Flash-Next is stricter: its Qwen Community License requires a separate licence for any such business, whatever its size.
Hosted prices in Alibaba Cloud's Singapore region run from US$0.03 per million input tokens for qwen3.7-flash on short prompts to US$2 input and US$6 output for qwen3.8-max, and new users get a million free tokens per model for 90 days there. For ASEAN teams, the practical appeal is choice: self-host an Apache model, pay per token in Singapore, or use the free Qwen Chat. The cost is keeping track of which licence and which version you are actually running.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
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 →
Alternatives to Qwen
More in LLMs & Chat
