Qwen3.8-27B

Qwen3.8-27B is Qwen's dense open-weight vision-language model for coding, professional work, research, and agent tasks, with a stable hosted API and practical self-hosting paths.

Language ModelOpen WeightsMultimodalLong ContextReasoning ModelQwen

QwenCloud hosted API or self-managed Transformers, vLLM, SGLang, and TokenSpeed inference

$0.50/1M input, $0.10/1M implicit-cache input, and $3/1M output tokens on QwenCloud; open weights have no provider token price

Commercial

🚀Function Overview

A dense multimodal Qwen model for teams that want official Apache-2.0 weights and local control without giving up a stable hosted model ID, published pricing, or managed tools.

Key Features

Official open checkpoint ID Qwen/Qwen3.8-27B
Stable hosted model ID qwen3.8-27b on QwenCloud
Dense 27-billion-parameter language model with a vision encoder
Text, image, and video input with text output
262,144-token native context extensible to 1,000,000 tokens
Thinking enabled by default but configurable per request
Low, medium, and xhigh reasoning effort, with xhigh documented as the default
Function calling, structured output, batches, web search, caching, and fine-tuning on QwenCloud
Official Transformers, vLLM, SGLang, and TokenSpeed compatibility
Apache License 2.0 open weights

Use Cases

•Local or self-hosted coding and agent workflows with image or video context
•Long-document, repository, and professional-work analysis
•Multimodal document, diagram, screen, and video understanding
•Hosted production use through QwenCloud when managed tools, quotas, and token pricing are preferable
•Controlled comparison with Qwen3.8-Max, Qwen3.6-27B, or other local and hosted models

⚙️Input Parameters

model

string

Use qwen3.8-27b on QwenCloud, Qwen/Qwen3.8-27B for supported model hubs, or the exact local model path served by your inference framework.

messages

array

Conversation messages containing text and supported image or video content.

enable_thinking

boolean

Enable or disable thinking where the selected API or serving framework supports this control.

reasoning_effort

string

Set reasoning depth to low, medium, or xhigh. Qwen documents xhigh as the default, so lower it for routine tasks when latency and token use matter.

preserve_thinking

boolean

Control whether supported serving paths retain reasoning context from earlier messages.

💡Usage Examples

Example 1

Input Parameters

{
  "model": "qwen3.8-27b",
  "messages": [
    {
      "role": "user",
      "content": "Review this architecture diagram and migration note, then identify the smallest safe validation sequence."
    }
  ],
  "enable_thinking": true,
  "reasoning_effort": "medium"
}

Output Results

A multimodal chat completion from the hosted or self-managed Qwen3.8-27B endpoint. Validate reasoning schema, latency, token use, and tool behavior on the selected serving path.

Quick Actions

Technical Specifications

Hardware Type: QwenCloud hosted API or self-managed Transformers, vLLM, SGLang, and TokenSpeed inference
Commercial Use: Supported
Pricing: $0.50/1M input, $0.10/1M implicit-cache input, and $3/1M output tokens on QwenCloud; open weights have no provider token price

Related Keywords

Qwen3.8-27Bqwen3.8-27bQwen 3.8 27B open weightsQwen3.8-27B pricingQwen3.8-27B context lengthQwen3.8-27B local deploymentQwen3.8-27B vLLMQwen3.8-27B SGLangQwen3.8-27B VRAMQwen3.8-27B reasoning effort

Related Models

Bielik 1.5B v3 Instruct

Bielik-1.5B-v3-Instruct is a generative text model featuring 1.6 billion parameters. It is result of collaboration between the open-science/open-souce project SpeakLeash and the High Performance Computing (HPC)

Cordia-A6 Text Generation Model

A model for generating text sequences based on input prompts and adjustable parameters.

Claude Sonnet 4

Claude Sonnet 4 is a significant upgrade to 3.7, delivering superior coding and reasoning while responding more precisely to your instructions