Alibaba Releases 2.4T Parameter Qwen3.8-Max, Open Weights Coming Next Week

Alibaba's Qwen team has officially launched its most powerful model to date, Qwen3.8-Max, with a total parameter count of 2.4T. It sets a new benchmark in autonomous programming and collaboration, and has topped the Text Arena. The team also announced that the weights for this model and Qwen3.8-27B will be open-sourced next week. This marks the first time Tongyi Qianwen has open-sourced a Max-tier model, and the second industry-wide open-source model exceeding 2T parameters following Kimi-K3, making it highly noteworthy for developers.

Confirmed

Key point: Alibaba released the Qwen3.8-Max model with a total of 2.4T (2 trillion) parameters.
Key point: The model sets new standards in autonomous programming and collaboration, and currently tops the Text Arena leaderboard.
Key point: It is officially confirmed that the weights for both Qwen3.8-Max and Qwen3.8-27B will be released and open-sourced next week.
Key point: Inference provider TokenSpeed announced Day-0 support for the model and is currently optimizing multi-node inference.

Unconfirmed

Key point: Author @cedricchee hopes the open-source model will adopt a more permissive MIT license in the future, though the exact licensing details are yet to be announced.

Why it matters

Key point: This is the first time Tongyi Qianwen has open-sourced a Max-tier model, significantly advancing its open-source ecosystem and providing the developer community with a more powerful foundational model option.
Key point: Open-source models with over 2T parameters remain rare in the industry; this move follows Kimi-K3 and further propels the open-source progress of ultra-large models.

2026-08-03 ~ 2026-08-03 · 12 related posts

Primary sources

Alibaba Launches 2.4T Parameter Qwen3.8-Max, Open Weights Coming Next Week — Alibaba_Qwen · 2026-08-03 10:15 (UTC+08:00)
Qwen3.8-Max Hits Text Arena: 2.4T Parameters Focused on Autonomous Coding — Alibaba_Qwen · 2026-08-03 10:27 (UTC+08:00)
Qwen Open-Sources First Max-Class Model, Exceeding 2T Parameters — cedric_chee · 2026-08-03 10:38 (UTC+08:00)

Alibaba's Qwen3.8-Max Released, TokenSpeed Announces Inference Optimization — zhyncs42 · 2026-08-03
[source] Alibaba Launches 2.4T Parameter Qwen3.8-Max, Open Weights Coming Next Week — Alibaba_Qwen · 2026-08-03
[source] Qwen3.8-Max Hits Text Arena: 2.4T Parameters Focused on Autonomous Coding — Alibaba_Qwen · 2026-08-03
[source] Qwen Open-Sources First Max-Class Model, Exceeding 2T Parameters — cedric_chee · 2026-08-03
Users Call for 200B Parameter Open-Source Release from Qwen — adrianscottcom · 2026-08-03
Alibaba launches Qwen3.8-Max at 2.4T parameters, excelling in autonomous coding — cephaloform · 2026-08-03
Qwen3.8-Max Hits #4 on Frontend Code Arena, Reshaping Cost-Performance Frontier — Scobleizer · 2026-08-03
Alibaba to Open-Source Qwen3.8-Max with 2.4T Parameters and Autonomous Coding — EugeneWare · 2026-08-03
Alibaba Announces Qwen3.8-Max: 2.4T Parameters and 1M Context Window — Scobleizer · 2026-08-03

3 near-duplicate retellings: yacineMTB · airesearch12 · 千问大模型