Qwen3.8

Alibaba Qwen🇨🇳 CNFrontier· Aug 5, 2026

TextReasoningVisionopen weightsAPI

Dense 27B vision-language model with native image and video understanding, adjustable thinking depth, and a focus on coding and long-horizon agentic tasks.

Summary of the model card

Context: 1M tokens
Price / 1M tok: free in · free out
License: apache-2.0
Modalities: text, image, video → text

7 variants

27B	27B	1M	$0.42/$3	OR
27B Free	27B	262K	free/free	OR
2.4T A95B	2.4T-A95B	1M	$2/$6	OR
27B	27B			HF
27B FP8	27B			HF
2.4T A95B	2.4T-A95B			HF
2.4T A95B FP8	2.4T-A95B			HF

Sources

HuggingFace OpenRouter Blog

Deutsch

Dichtes 27B Modell mit nativem Bild- und Videoverständnis, dessen Denktiefe über reasoning_effort steuerbar ist. Es ist für Coding und mehrstufige agentische Aufgaben ausgelegt.

Provenance

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