Qwen3.8-Max: Alibaba's Open-Weight Flagship, Explained

Alibaba's Qwen3.8-Max is a forthcoming 2.4 trillion-parameter large language model set to be released with open weights, making its advanced capabilities acces…

Qwen3.8-Max Reaches General Availability

At 2.4 trillion total parameters with roughly 95 billion active per forward pass, this is the largest model Alibaba has shipped — 1M-token context, three reasoning level…

Qwen3.8-Max Previewed at WAIC Shanghai

The preview endpoint qwen3.8-max-preview exposed the same 2.4T-parameter architecture two weeks before GA, arriving days after rival lab Moonshot's Kimi K3 open-weight l…

Qwen3.7-Max Launches as a Closed-Weight Flagship

Reportedly scored 56.6 on the Artificial Analysis Intelligence Index v4.0 — the highest any Chinese-developed model had reached at the time. API-only, 1M-token context,…

Qwen3.6 Ships an Open-Weight MoE Model

The open-weight Qwen3.6-35B-A3B sparse Mixture-of-Experts model shipped under Apache 2.0 — 35B total parameters, roughly 3B active, genuinely self-hostable on a single h…

Qwen3.5 Adds Computer-Use Agent Capability

A 397B-A17B Mixture-of-Experts model, Apache 2.0 for the open variant, reportedly adding desktop- and mobile-operation agent capability beyond text generation.

Qwen3 Establishes Apache 2.0 as the Standard

Dense models from 0.6B to 32B parameters plus MoE variants at 30B-A3B and 235B-A22B, trained on a reported 36 trillion tokens across 119 languages. From this release for…

QwQ Brings Dedicated Reasoning Models to Qwen

QwQ-32B-Preview shipped November 2024; full QwQ-32B followed March 2025, both Apache 2.0 at a modest 32B-parameter size, arriving as OpenAI's o1 popularized dedicated re…

Qwen2.5 Expands the Family

Broadened Qwen2's size range with mixed licensing across sizes; Qwen2.5-Omni followed in early 2025 with multimodal input/output including voice.

Qwen2 Adds Mixture-of-Experts and Multimodal Variants

Four dense sizes (0.5B–72B) plus Alibaba's first Qwen-branded sparse MoE model, Qwen2-57B-A14B, under Apache 2.0. Qwen2-Audio and Qwen2-VL followed later in 2024.

Qwen's First Open-Weight Releases

Initial 7B weights released August 2023; 72B and 1.8B models followed by December, alongside Qwen-VL, a vision-language variant.

Tongyi Qianwen Is Announced

Alibaba's original bilingual (Chinese/English) foundation model line is announced in beta, developed under Alibaba Cloud with research roots in Alibaba's DAMO Academy.