Qwen3.8 Flash

Alibaba Qwen🇨🇳 CNFrontier· Aug 24, 2026

TextReasoningVisionopen weightsAPI

Multimodal mixture-of-experts model aimed at long-context agentic coding and document workflows, with open weights for the Flash-Next preview. It combines Gated DeltaNet with Qwen Sparse Attention for efficient inference over up to 1M tokens.

Summary of the model card

Context: 1M tokens
Price / 1M tok: $0.15 in · $0.47 out
License: other
Modalities: text, image, video → text

3 variants

Standard			HF
FP8			HF
Standard	1M	$0.15/$0.47	OR

Sources

HuggingFace OpenRouter Blog

Deutsch

Multimodales Sprachmodell mit Mixture-of-Experts-Architektur für langen Kontext und agentisches Coding mit bis zu 1M Tokens. Die offenen Gewichte stammen aus der Preview Qwen3.8-Flash-Next, deren Architektur Gated DeltaNet mit Qwen Sparse Attention kombiniert.

Provenance

Detected: Sep 27, 2026
First source: openrouter
Description: Summary of the model card
Editorially reviewed: —