Qwen3-Next-80B-A3B-Instruct

The non-thinking instruction release of Qwen3 Next.

Practical use

Practical guidance. Use the instruction checkpoint when your application expects direct responses. Keep the parser independent of reasoning markers so it does not assume output conventions from the separate Thinking checkpoint.

Specifications & access

Qwen3-Next-80B-A3B-Instruct specifications
Detail	Value and conditions
Access	Downloadable
Model identifiers	`Qwen/Qwen3-Next-80B-A3B-Instruct`
Architecture	80 billion total and three billion active parameters; hybrid gated delta and gated attention layers.Qwen3-Next-80B-A3B-Instruct model card
Context	256,000-token class context.Qwen3-Next-80B-A3B-Instruct model card
Licence	Apache 2.0.Qwen3-Next-80B-A3B-Instruct model card

Related models

Qwen3-Next-80B-A3B-Thinking
Alibaba · Qwen3 Next
The reasoning-trained release of Qwen3 Next.
Launch date not confirmedDownloadable