# Alibaba expands Qwen3.8 AI offerings and agent ecosystem > Live situation record from CLSTR: https://clstr.news/situations/alibaba-cloud-qwen-model-rollout > Updated: 2026-09-17T23:15:12.000Z. Sources: 129. Developments: 18. Following the July 2024 unveiling of the 2.4-trillion-parameter Qwen 3 model, Alibaba introduced the Qwen 3.8-Max flagship in early August 2024. This multimodal mixture-of-experts system activates approximately 95 billion parameters per request, supports text, images, video, and up to one million tokens, and achieved benchmark results comparable to Anthropic’s Claude Fable 5. Alibaba released the model’s weights publicly via Model Studio the week of 10 August, which contributed to a roughly 7% rise in its Hong Kong shares. In August 2026, the company expanded its ecosystem by launching the Qwen open platform, allowing third-party developers to create AI agents for mobile, PC, and AI glasses. Initial partners include SF Express and Lenovo. On 13 August 2026, the Qwen team released the open weights for the Qwen 3.8 model family under the Apache 2.0 license. This included the Qwen 3.8-27B, a multimodal model designed for consumer-grade hardware with a native context window of 262,144 tokens, extendable to 1 million via YaRN. By mid-August 2026, Alibaba’s Qwen model family reached a milestone of over 3 billion global downloads within six months, though Hugging Face reported approximately 2.045 billion. On 19 August 2026, OrcaRouter released an uncensored version of the Qwen 3.8-27B model optimized for Apple Silicon Macs, featuring an “Aggressive” variant to provide more direct responses. On 25 August 2026, Alibaba released Qwen3.8-Flash-Next, an open-weight preview of its upcoming Qwen4 architecture. On 26 August 2026, Alibaba launched the international public beta of QwenWork, an all-in-one AI agent platform for individuals and enterprises in Asia, the Middle East, and Latin America. On 2 September 2026, Alibaba announced the Qwen 3.8 series, emphasizing the flagship model’s 2.4 trillion parameters and its availability as open weights for self-hosting and fine-tuning. On 17 September 2026, controversy arose when the Federal Register, managed by the U.S. ## Claims - Alibaba claims its Qwen AI models have surpassed 3 billion global downloads. (disputed by 11 sources) - Alibaba states its Qwen ecosystem has produced over 300,000 derivative models. (disputed by 8 sources) - Hugging Face reports approximately 2.045 billion Qwen downloads, a lower figure than Alibaba's claim. (disputed by 3 sources) - Hugging Face counts 151,448 Qwen-based derivatives on its platform. (disputed by 2 sources) - Qwen3.8‑Max can process up to one million tokens per operation. (corroborated by 37 sources) - Alibaba will release Qwen3.8‑Max weights for public download in the week starting 10 August. (corroborated by 37 sources) - Alibaba's shares rose about 7 % in Hong Kong after the Qwen3.8‑Max announcement. (corroborated by 36 sources) - Qwen3.8‑Max completed an autonomous software‑engineering project in 16 days in internal testing. (corroborated by 36 sources) - Qwen3.8‑Max’s benchmark performance is comparable to Anthropic’s Claude Fable 5. (corroborated by 30 sources) - On the Arena AI leaderboard Qwen3.8-Max ranks second globally for image‑analysis tasks, behind Anthropic’s Claude Fable 5. (corroborated by 25 sources) ## Timeline ### 2026-09-17: Alibaba AI model used by U.S. government website The U.S. Federal Register website removed an AI search tool powered by Alibaba’s Qwen model following social media scrutiny and recent FBI allegations regarding the copying of U.S. technology. 4 sources. https://clstr.news/cluster/alibaba-ai-model-used-by-us-government-website ### 2026-09-07: Alibaba releases Qwen3. 8-Flash open-weight multimodal model Alibaba has unveiled Qwen3. 8-Flash, an open-weight multimodal MoE model optimized for coding and high-volume tasks with efficient parameter activation. 2 sources. https://clstr.news/cluster/alibaba-releases-qwen3-8-flash-open-weight-multimodal-model ### 2026-09-03: Chinese AI models emerge as competitive, low-cost alternatives to US peers Chinese AI models like Alibaba's Qwen are becoming competitive, low-cost alternatives to US models, offering advanced coding and visual analysis capabilities while sparking discussions on security. 3 sources. https://clstr.news/cluster/chinese-ai-models-emerge-as-competitive-low-cost-alternatives-to-us-peers ### 2026-09-02: Alibaba unveils Qwen 3.8 AI model series with 2.4T parameters Alibaba has unveiled its Qwen 3.8 AI model series, featuring a 2.4 trillion parameter flagship with open weights and a high-performing, cost-effective 0902-version optimized for coding and office tasks. 4 sources. https://clstr.news/cluster/alibaba-unveils-qwen-38-ai-model-series-with-24t-parameters ### 2026-09-01: Alibaba launches QwenWork International AI agent platform Alibaba has released the international beta of QwenWork, an AI agent platform that automates desktop and web-based workplace tasks using natural-language instructions. 3 sources. https://clstr.news/cluster/alibaba-launches-qwenwork-international-ai-agent-platform-1 ### 2026-08-29: Wan 3.0 AI video model integrates with Qwen and SkyProduction Alibaba’s Wan 3.0 video model is being integrated into platforms like Qwen and SkyProduction, introducing multi-agent collaboration and advanced multimodal controls for professional AI video production. 2 sources. https://clstr.news/cluster/wan-30-ai-video-model-integrates-with-qwen-and-skyproduction ### 2026-08-26: Alibaba launches QwenWork international AI agent platform Alibaba has released the international public beta of QwenWork, an AI agent platform that automates complex workplace tasks through natural language commands and multimodal generation. 5 sources. https://clstr.news/cluster/alibaba-launches-qwenwork-international-ai-agent-platform ### 2026-08-22: Qwen 3.8 27B model reaches one million downloads The Qwen 3.8 27B open-weights model has reached one million downloads in two days, earning high intelligence scores and popularity for running complex reasoning tasks on consumer hardware. 2 sources. https://clstr.news/cluster/qwen-38-27b-model-reaches-one-million-downloads ### 2026-08-19: OrcaRouter releases uncensored Qwen AI model and developer support program OrcaRouter has launched an uncensored version of Alibaba’s Qwen 3.8-27B AI model for local use on Mac hardware and introduced a new program to provide revenue sharing for open-source developers. 3 sources. https://clstr.news/cluster/orcarouter-releases-uncensored-qwen-ai-model-and-developer-support-program ### 2026-08-15: Chinese AI labs lead in open-source model scale, Hugging Face report A Hugging Face report shows Chinese AI labs are leading in frontier-scale open models, with Alibaba’s Qwen becoming a dominant community standard for derivative development. 4 sources. https://clstr.news/cluster/chinese-ai-labs-lead-in-open-source-model-scale-hugging-face-report ### 2026-08-15: Alibaba's Qwen AI models surpass 3 billion downloads Alibaba's Qwen AI models have reached over 3 billion downloads, outpacing open-weight models from Meta and Google to become a leading foundation in the global AI ecosystem. 24 sources. https://clstr.news/cluster/alibaba-strengthens-ai-position-with-qwen-model-downloads ### 2026-08-13: Alibaba releases open weights for Qwen3. 8 AI model family Alibaba has released open weights for its Qwen3. 8 AI models, including the multimodal 27B version for local use and the massive 2.4 trillion parameter Qwen3. 8-Max for enterprise workloads. 10 sources. https://clstr.news/cluster/motif-technologies-releases-motif-3-ai-model-as-white-house-expands-safety-oversight ### 2026-08-10: Alibaba launches Qwen platform as 01.AI shifts focus to enterprise AI Alibaba launched its Qwen open platform to build an AI agent ecosystem, while 01.AI announced it will phase out consumer services to focus on enterprise AI solutions. 3 sources. https://clstr.news/cluster/alibaba-launches-qwen-platform-as-01ai-shifts-focus-to-enterprise-ai ### 2026-08-03: Alibaba launches public beta of QwenWork enterprise AI office tool Alibaba has opened public beta of QwenWork, an enterprise AI office suite powered by the Qwen 3.8 model, with web and PC access and upcoming DingTalk integration, offering free and paid subscription tiers. 3 sources. https://clstr.news/cluster/alibaba-launches-public-beta-of-qwenwork-enterprise-ai-office-tool ### 2026-08-03: Alibaba launches Qwen3.8‑Max, its most powerful AI model Alibaba released Qwen3.8‑Max, a 2.4 trillion‑parameter AI model that ranks second in visual benchmarks, completed a 16‑day autonomous coding test, and will have open weights released on Aug 10; shares jumped 7 68 sources. https://clstr.news/cluster/alibaba-launches-24trillionparameter-qwen38max-ai-model ### 2026-07-19: Alibaba unveils Qwen 3.8 large language model Alibaba introduced Qwen 3.8, a 2.4 trillion‑parameter LLM claiming to trail only Anthropic’s Fable 5, but without benchmark data, as Chinese AI rivals push the competition forward. 3 sources. https://clstr.news/cluster/alibaba-unveils-qwen-38-large-language-model ### 2026-07-19: Alibaba Cloud launches Token Plan for Qwen3.8-Max AI model Alibaba Cloud launches a Token Plan personal subscription, priced from 39 CNY/month, to give early access to the high‑performance Qwen3.8‑Max AI model. 5 sources. https://clstr.news/cluster/alibaba-cloud-launches-token-plan-for-qwen38-max-ai-model ### 2026-06-29: New AI Model Deployment Guides Released for Multiple Large Language Models AI developers released zero‑config install guides for OmniVoice, chandra‑ocr‑2, Qwen3‑VL‑32B‑Instruct, Qwen3.6‑27B‑NVFP4, Qwen3.5‑4B, and DeepSeek‑OCR‑2, enabling quick local deployment on Windows/WSL. 6 sources. https://clstr.news/cluster/alibaba-clouds-qwen3-ai-models-receive-new-local-deployment-guides --- Cite as: Alibaba expands Qwen3.8 AI offerings and agent ecosystem. CLSTR, https://clstr.news/situations/alibaba-cloud-qwen-model-rollout