Qwen3.8 27B

Qwen3.8-27B is a dense 27-billion-parameter vision-language model built on the Qwen3.5 architecture. It combines a causal language model with a vision encoder, positioning it for coding, professional tasks, research, and multi-step agentic workflows. Its native context accommodates very long inputs, while RoPE scaling can extend the working window toward one million tokens for unusually large documents, codebases, or media sequences.

The model is designed for practical work that mixes visual and textual information, including documents, STEM diagrams, and video, while configurable reasoning and retained thinking context help maintain continuity during complex tasks. It supports tool use and reasoning, and downloadable GGUF and MLX versions make it suitable for local experimentation or deployment. The strongest fit is therefore multimodal analysis, code assistance, and extended agentic projects where long context and sustained reasoning matter more than compact size.

Cloudflare Workers AI@cf/qwen/qwen3.8-27bqwen

Open provider page Provider docs

Quick Info

Provider: Cloudflare Workers AI
Model key: @cf/qwen/qwen3.8-27b
Release date: Aug 14, 2026
Last updated: Aug 14, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.45
Output token cost: $3.20

Limits

Output tokens: 262,144 tokens
Context window: 262,144 tokens

Transparent token rates

Compare Qwen3.8 27B pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Qwen3.8 27B

Cloudflare Workers AI

Coverage

Rost Glukhov's Medium analysis frames Qwen3.8-27B as the most consequential local-AI release of 2026 for developers who want to own, run, and customize their own model. The article confirms both Qwen3.8-Max and Qwen3.8-27B open weights are now available on Hugging Face (Qwen/Qwen3.8-2.4T-A95B and Qwen/Qwen3.8-27B) and The piece leans on the official Qwen account's announcement that open weights for Qwen3.8-Max and Qwen3.8-27B were being released, positioning the 27B as the practical counterpart to the flagship. Technical detail is light and largely interpretive; the article offers community sentiment and developer-adoption framing r

Cloudflare Workers AI

CoverageBenchmark

Alibaba's Qwen team released Qwen3.8-27B on August 14, 2026, shipping a dense 27.78-billion-parameter multimodal checkpoint that accepts text, images, and video under an Apache 2.0 license with a native 262,144-token context window. This is the same base model served as @cf/qwen/qwen3.8-27b on Cloudflare Workers AI, ma Qwen reports substantial improvements over Qwen3.6-27B in agentic coding, computer use, and vision-language work without increasing decoder size: Terminal-Bench 2.1 rose from 63.4 to 73.0, DeepSWE 1.1 from 13.3 to 42.2, OSWorld-Verified from 63.9 to 84.3, and SWE-MM from 25.7 to 38.6. Kingy.ai's launch-day review flags

Cloudflare Workers AI

CoverageAnalysis

The Local AI Zone technical analysis confirms Alibaba's Tongyi Lab released Qwen3.8-27B on August 14, 2026, 15:00 UTC under Apache 2.0, hosted on Hugging Face. The architecture breakdown specifies 27,781,427,952 parameters across 64 Transformer blocks with a 5,120 hidden dimension, 24 query heads / 4 KV heads at head d A signature architectural choice is the 3:1 hybrid attention ratio: 48 Gated DeltaNet (linear attention) layers and 16 Gated Full Attention layers, which reduces computational complexity from O(n²) to O(n) for 75% of layers while preserving long-context quality. An attention output gate fuses the q_proj at 12,288×5,120