< Back to situations

Get alerts on this situation

We’ll email you as it develops, and you can follow the whole thread from day one.

Unsubscribe anytime.

[SITUATION] · [ACTIVE]

5 clusters · 51 sources · 35 days · First seen about 1 month ago · Last updated about 11 hours ago

Categories: TECHNOLOGY

Alibaba launches 2.4T Qwen3.8‑Max model

Entities: Alibaba Group Holding Ltd · Qwen3.8-Max · Qwen3.8‑Max · Alibaba Group Holding · QwenWork

Overview

After June’s zero‑configuration deployment guides and the July announcement of the 2.4‑trillion‑parameter Qwen 3.8 model, Alibaba expanded its AI portfolio in early August with the public beta of QwenWork, a SaaS office platform that embeds Qwen 3.8 and offers free and paid plans for individuals and enterprises.

On 3 August 2026 the company unveiled Qwen3.8‑Max, a multimodal successor that processes text, images, video and up to one million tokens per request. Built on a mixture‑of‑experts architecture, it activates roughly 95 billion parameters per query, lowering latency and compute cost. In internal testing the model completed an autonomous software‑engineering project in 16 days and achieved benchmark scores comparable to Anthropic’s Claude Fable 5, ranking second worldwide on the Arena AI leaderboard for image‑analysis and becoming the highest‑ranking Chinese model for text tasks.

Alibaba announced that the model’s weights will be publicly downloadable the week of 10 August via Alibaba Cloud’s Model Studio, and it set token usage pricing at US$2 per million input tokens and US$6 per million output tokens. The launch lifted Alibaba’s Hong Kong‑listed shares by 5‑7 % and the firm plans to extend its shift from on‑premise guides to cloud‑based, token‑priced subscriptions and a full SaaS AI suite built around the Qwen 3.8 family.

Claims

What the coverage asserts, and how well corroborated each claim is across sources.

[● 25 SOURCES] Alibaba released the Qwen3.8-Max model with 2.4 trillion parameters. (Alibaba)
[● 25 SOURCES] Qwen3.8-Max can process text, images, video and up to one million tokens per operation. (Alibaba)
[● 25 SOURCES] The model uses a mixture‑of‑experts architecture that activates about 95 billion parameters per request. (Alibaba)
[● 25 SOURCES] On the Arena AI leaderboard Qwen3.8-Max ranks second globally for image‑analysis tasks, behind Anthropic’s Claude Fable 5. (Alibaba)
[● 25 SOURCES] Qwen3.8-Max’s benchmark performance is comparable to Anthropic’s Claude Fable 5. (Alibaba)
[● 25 SOURCES] Alibaba’s shares rose between 5 % and 7 % on the Hong Kong exchange after the model’s announcement. (Alibaba)
[● 25 SOURCES] Alibaba will release the Qwen3.8-Max model weights for public download in the week starting Aug 10 via Alibaba Cloud Model Studio. (Alibaba)
[● 25 SOURCES] During internal testing Qwen3.8-Max completed an autonomous software‑engineering project in 16 days. (Alibaba)

Timeline

about 11 hours ago

[TECHNOLOGY] 2 sources
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.
about 15 hours ago

[TECHNOLOGY] 37 sources
Alibaba launches Qwen3.8-Max, 2.4 trillion‑parameter AI model
Alibaba released Qwen3.8‑Max, a 2.4 trillion‑parameter AI model with mixture‑of‑experts architecture, open‑weight release on Aug 10, and shares up 5‑7% in Hong Kong.
15 days ago

[TECHNOLOGY] 3 sources
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.
15 days ago

[TECHNOLOGY] 5 sources
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.
about 1 month ago

[TECHNOLOGY] 6 sources
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.

Sources

36kr.com · assodigitale.it · brandsit.pl · cash.ch · cnmo.com · computerworld.in · cryptobriefing.com · drweb.de · dunya.com · eco.sapo.pt · edigest.hk · elektronikpraxis.de · emissoradasbeiras.pt · gigazine.net · hoshiimo.co.jp · hu.euronews.com · infomoney.com.br · infoquest.co.th · isd.sorbonneonu.fr · it-boltwise.de · ithome.com · jornaldenegocios.pt · leiphone.com · medmol.es · mirador.lu · mobil.zeit.de · mobile.valor.com.br · news.cnyes.com · newsable.asianetnews.com · noticiasdemalaga.es · paparazzi.com.ar · pro.hwupgrade.it · radiolideranca103.com.br · ravalsolidari.org · sgsme.sg · slashgear.jp · srnnews.com · sudefuturi.it · the-decoder.com · thenewstack.io · theprobatedepartment.co.uk · tmtpost.com · torresdelrio.es · trendingtopics.at · volksfreund.de · wdez.com · weekly.ascii.jp · weel.co.jp

This summary has been updated 5 times: see revision history