AITOPAI热报
Models

Qwen发布3.8 Max(2.4T)与27B开源模型，主打编程与协作

[AINews] Qwen 3.8 Max(2.4T) and 27B, new open weights models for Coding and Cowork

Official accountLatent Space (swyx)
Blog/Media
Read the original
Why we picked it

Qwen这次一口气放了2.4T和27B两个开源权重模型，专攻编程和协作用途，想自己部署或微调的直接看就行。

Qwen发布了两个新开源模型：Qwen 3.8 Max，参数规模2.

Body

4T，以及27B版本。两个模型都面向编程和协作场景，权重已经开放。

开发者可以基于这些权重进行本地部署或微调。这延续了Qwen在开源模型上的发布节奏。

Original · Latent Space (swyx)

[AINews] Qwen 3.8 Max(2.4T) and 27B, new open weights models for Coding and Cowork

Qwen is so back!

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Models" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Qwen, Qwen 3.8 Max, 开源模型, 编程助手.

Source attribution: curated from Latent Space (swyx).

#Qwen#Qwen 3.8 Max#开源模型#编程助手

Keep reading

Recent stories with the same tag or section
ProductsChina to ASEAN00:16
Chinese contentIOC主席Kirsty Coventry参观Alibaba达喀尔青奥会科技展台

Alibaba Cloud在达喀尔2026青奥会现场摆了个AI展台，访客能用Wan模型玩互动视频亭，还有Qwen驱动的聊天助手。中国云厂商把生成式AI直接搬进非洲体育赛事，想看Qwen实际落地场景的可以关注。

#Alibaba Cloud#Qwen#Wan#达喀尔2026青奥会
Official account阿里云 Alibaba Cloud@alibaba_cloudSource
PapersBuilders15:56
Chinese content论文揭示微调引发人格泛化遵循跨模型共享的线性规律

微调一个领域会意外改变模型在其他领域的行为，这篇论文给出了可预测的规律：域内外人格率呈线性关系，Qwen 和 Llama 都成立。做微调的开发者能直接用上——只调后几层可以在不牺牲任务效果的情况下减少不安全行为迁移。

#微调#emergent misalignment#Qwen#Llama
arXiv cs.LG@Olina Mukherjee et al. (4)Source
ProductsBuildersOfficial10:05
Chinese contentAlibaba 千问 Office 2.0 上线团队智能体空间与自动任务

Alibaba 给千问 Office 加了团队共享工作区，智能体直接挂在钉钉群和文档里干活，条件一满足就自动跑任务。和 Microsoft 365 Copilot 比，它吃的是钉钉这个入口，国内用钉钉办公的团队可以关注一下，看看能不能省掉一部分流程自动化配置。

#千问#Qwen#DingTalk#智能体
Primary sourcePandaily@[email protected] (Pandaily)Source
ModelsBuildersMulti-sourceSelected00:29
Ant open-sources LemonDecider, a 4B decision model beating Jev 1.14

Decision-specific micro-models are shifting routing and gating workloads away from large general LLMs, cutting both latency and cost in agent pipelines. Developers building multi-step agent workflows or RAG evaluators can drop LemonDecider in as a fast classifier, and its open training data makes custom fine-tuning feasible. Compared with general models like Qwen3.5-4B or Jev 1.14, it trades generality for a measurable benchmark lead and millisecond-level inference.

#LemonDecider#Ant Group#Qwen#open-source models
Event
AIGCLINK@aigclinkSource