Qwen

• June 22, 2026: Qwen Cloud model releases, Qwen3.7-Max, Qwen3.7-Plus, Qwen Cloud pricing, Qwen Studio, Qwen3 sources, and Hugging Face Qwen were rechecked. The newest official Qwen Cloud changelog entry remains qwen3.7-max-2026-06-08, listed on June 10 with visual-modal understanding added versus the May 20 Max snapshot, while the live qwen3.7-max marketplace page still describes public experimentation as text-only. Verify modality on the exact endpoint before building around visual input.
• June 22, 2026: The public qwen.ai surface is now best described as Qwen Studio, not just Qwen Chat. Qwen’s official research page also added the Qwen-Robot Suite on June 16, including Qwen-RobotNav, Qwen-RobotManip, and Qwen-RobotWorld.
• June 22, 2026: Pricing docs still list qwen3.7-max at $2.50/M input and $7.50/M output, with the 50% promo page expiring June 22. qwen3.7-plus list rates remain $0.40/$1.60 up to 256K and $1.20/$4.80 from 256K-1M, while its model page shows 20% off display pricing without an official expiry found.
• June 15, 2026: Qwen Cloud model releases, Qwen3.7-Max, Qwen3.7-Plus, Qwen Cloud pricing, the Qwen3.7-Max promo page, Qwen3 sources, and Hugging Face Qwen were rechecked again. No material change was found versus the June 14 refresh.
• June 6, 2026: Qwen Cloud model releases, Qwen3.7-Max, Qwen3.7-Plus, Qwen Cloud pricing, and Qwen3 sources were rechecked while refreshing Mistral AI vs Qwen. The buyer split is now explicit: Mistral is the EU/open-model/vendor-platform lane; Qwen is the Alibaba/Qwen Cloud, multilingual, qwen3.7-max, qwen3.7-plus, and Qwen3 open-weight lane.
• May 27, 2026: Alibaba Cloud used its first international Qwen Conference to push Qwen as an agent-cloud platform. The buyer signal is that Qwen is moving beyond model-family benchmarking into Qwen Cloud, Skills, infrastructure upgrades, and enterprise agent tooling.
• May 27, 2026: Qwen Cloud’s model-release changelog lists qwen3.7-plus / qwen3.7-plus-2026-05-26 as a multimodal interactive hybrid-agent model for screen/GUI perception, code generation from visual references, tool use, productivity workflows, and end-to-end mobile-app navigation.
• May 22, 2026: Qwen Cloud’s model-release changelog added qwen3.5-livetranslate-flash-realtime and its 2026-05-19 snapshot. It is the newest specialty Qwen release AiPedia found in official sources, aimed at real-time multilingual audio/video translation.
• May 21, 2026: Qwen Cloud listed qwen3.7-max / qwen3.7-max-2026-05-20 as the next-generation flagship in the Qwen Max series. The official model page shows text input/output, thinking enabled by default, a 1M context window, 991K max input, 65K max output, and list pricing of $2.50/M input and $7.50/M output.
• May 13, 2026: AiPedia refreshed this page against official Qwen, Alibaba Cloud Model Studio, Hugging Face, and the latest Qwen ecosystem coverage. Model Studio International pricing now lists qwen-max at $1.20/M input (0-32K) and $6.00/M output (down from the May 10 list of $1.60/$6.40), with a new Qwen-Flash tier at $0.10/M input and $0.40/M output. Qwen-Turbo is no longer receiving updates; Qwen-Flash is the recommended replacement.
• May 11, 2026: Alibaba Qwen and Taobao launched a co-built agentic shopping experience. The integration is the highest-profile production deployment of Qwen agentic capabilities to date, pushing the family from developer-facing model lineup into a consumer-scale commerce surface that touches hundreds of millions of users.
• April 10, 2026: Vidu Shengshu, the Alibaba-affiliated video-model studio, raised fresh funding. Reinforces that Alibaba’s AI bet now spans Qwen text/code, Qwen-VL, image, video, and embodied stacks, not just the chat model family.
• April 16, 2026: Third-party coverage reported a Qwen3.6-35B-A3B sparse MoE release. AiPedia is tracking it as a market signal, but this evergreen page keeps the official Qwen3 open-weight line as the buyer-facing baseline until primary source support is clear.
• April 30, 2026: Alibaba-linked Metis showed an 8B Qwen3-VL-based agent can improve by calling tools less. The HDPO-trained model reduces blind tool calls from 98% to 2% in the project reports, making tool abstention a useful Qwen ecosystem signal.
• April 19, 2026: Alibaba Amap debuts first embodied robot at Beijing Humanoid Robot Half Marathon. Quadruped from Amap’s new embodied-intelligence division, powered by Alibaba’s ABot-World model (leads AGIbot World Challenge and World Arena benchmarks). Moves Alibaba from Qwen-as-foundation into first-party robotics alongside the model family.

Pick Qwen if you need open-weight models with multilingual reach. Apache 2.0 Qwen3 releases give real commercial flexibility. The official Qwen3 release lists 119-language coverage and model sizes from 0.6B to 235B MoE, making Qwen a strong candidate for multilingual products, local experiments, and custom hosted deployments.

Skip it if you want a polished consumer chat product or strict Western data residency. Qwen Studio is useful for testing, but it is not ChatGPT-grade as a general consumer workspace. Alibaba Cloud is a Chinese provider, which matters for regulated enterprise buyers. Competing open-weight families like DeepSeek may be stronger on specific reasoning or cost benchmarks.

Who uses which surface: Qwen Studio for quick tests, Hugging Face or ModelScope downloads for self-hosters, Alibaba Cloud Model Studio for hosted API use, and third-party gateways only after checking their separate pricing and model availability.

Qwen3.7-Max, Qwen3.7-Plus, Qwen Cloud pricing, and model-release rows above were verified on 2026-06-22. Older qwen-max examples retain their own source dates in price history. See Sources.

• Consumer chat product is minimal. qwen.ai is functional for testing but lacks ChatGPT-grade onboarding, memory, or ecosystem.
• Data residency on Alibaba Cloud. Enterprise buyers in regulated industries need to evaluate the Chinese-cloud posture. Self-hosting the Apache 2.0 weights is the workaround.
• Thin moat on open-weight leaderboard. DeepSeek, Kimi, GLM, and Qwen all iterate monthly. Leadership positions shift fast.
• English documentation lag. Official docs translate from Chinese first. Some resources trail the Chinese original by weeks.
• Vision models lag the strongest closed models. Qwen-VL and Qwen3.5-Omni are capable but trail the strongest closed vision models on independent evaluations.
• Hosted API rate limits vary by region. Alibaba Cloud tier and regional load affect throughput. Production deployments should load-test.
• Pricing is model-specific. Alibaba Cloud Model Studio tables change by model, mode, free quota, context, and batch eligibility.
• Changelog and marketplace wording can diverge. The June 10 changelog says the June 8 Max snapshot adds visual-modal understanding, while the live qwen3.7-max marketplace page still describes a text-only public interface. Verify the exact route before promising visual input.
• Latest does not mean open weight. Qwen3.7-Max is a hosted Qwen Cloud flagship route. The Apache 2.0 open-weight buyer case still rests on the official Qwen3 checkpoints.
• Promos can distort cost comparisons. Qwen Cloud showed a 50% Qwen3.7-Max promotional rate during this refresh; compare on list price unless you are buying during the promo window.
• Pricing pages can disagree. The representative pricing docs and model pages are both official, but model pages can show temporary display discounts. Recheck the exact page before publishing a cost comparison.
• Responses API has separate retention behavior. Qwen Cloud says normal API inputs and outputs are not used for training, while linked conversation context for the Responses API is stored for 7 days.

This page was produced by the aipedia.wiki editorial pipeline, an automated system that ingests vendor documentation, verifies pricing and model details against primary sources, and generates the editorial analysis you are reading. No individual human wrote this review. Scoring follows the four-dimension rubric at /about/scoring/ (Utility, Value, Moat, Longevity; unweighted average).

Last verified 2026-06-22 against Qwen Cloud model releases, the Qwen3.7-Max model page, Qwen Cloud pricing docs, the Qwen3.7-Max promotion page, Qwen official site, Qwen3 blog, Hugging Face Qwen, Qwen Studio and Qwen research pages, current Qwen Conference coverage, Qwen-Taobao coverage, and tracked Qwen3.6-35B-A3B coverage.

Is Qwen open source? Partly. The official Qwen3 open-weight line ships under Apache 2.0 on Hugging Face and ModelScope, covering sizes from 0.6B to 235B MoE. Download, self-host, fine-tune, and deploy commercially under that license, but verify the exact model because not every Qwen-branded surface is open.

What is the main Qwen3 open-weight release? The official Qwen3 release includes two MoE models, Qwen3-235B-A22B and Qwen3-30B-A3B, plus six dense models from 0.6B through 32B. Qwen says the line supports hybrid thinking modes, 119 languages and dialects, and agentic/coding improvements.

What is the latest Qwen model? As of this refresh on June 22, 2026, the latest official Qwen Cloud changelog entry AiPedia found is qwen3.7-max-2026-06-08, listed on June 10 as a Max snapshot with visual-modal understanding added versus the May 20 snapshot. The live qwen3.7-max marketplace page still describes the public model page as text input/output, so buyers should verify the exact route before assuming visual input. The current Plus multimodal/GUI agent lane remains qwen3.7-plus-2026-05-26. None of this changes the buyer-facing fact that the main open-weight line is still Qwen3.

How does Qwen3 compare to Claude? Qwen is more compelling when you need open weights and self-hosting. Claude is usually stronger when you want a polished paid assistant or API for English writing, long-document work, and managed enterprise workflows.

Can I run Qwen locally? Yes. Official Qwen3 sizes start at 0.6B and scale up to 235B MoE. Practical hardware depends on model size, quantization, context length, throughput targets, and serving stack.

• Qwen official site: Qwen Studio and model-family surface
• Qwen3 official blog: architecture, open-weight models, training, 119-language coverage
• Qwen Cloud model releases: qwen3.7-max June 8 snapshot, qwen3.7-plus, and Qwen3.5 LiveTranslate release dates
• Qwen3.7-Max model page: model alias, context, built-in tools, and current marketplace pricing display
• Qwen Cloud pricing docs: pay-as-you-go text model pricing, including Qwen3.7-Max and Qwen3.7-Plus representative rates
• Qwen3.7-Max promotion: temporary 50% discount through June 22, 2026
• Alibaba Cloud Qwen Conference coverage: Qwen Cloud, Skills, infrastructure upgrades, and JVS Agent Suite
• Alibaba Cloud Model Studio pricing: current hosted rates
• Hugging Face Qwen: open-weight model downloads

• Category: AI Chatbots · AI Coding
• Comparisons: DeepSeek vs Qwen · Mistral AI vs Qwen