Qwen3.8 27B API

Qwen3.8 27B reads text, images and video and answers in text, from $0.180 per million input tokens.

1M token context
27.8B dense parameters
FP8 precision served
Apache 2.0 license

from openai import OpenAI

client = OpenAI(
    base_url="https://api.modelsheep.com",
    api_key=API_KEY_MODELSHEEP,
)

response = client.chat.completions.create(
    model="Qwen/Qwen3.8-27B-FP8",
    messages=[
        {
            "role": "user",
            "content": "Why do sheep count people?"
        }
    ],
)

print(response.choices[0].message.content)

How to use the Qwen3.8 27B API

1
Create a ModelSheep account and copy an API key from the dashboard.
2
Point the OpenAI SDK at the base url https://api.modelsheep.com and set the model to Qwen/Qwen3.8-27B-FP8.
3
Call /chat/completions with your messages and set reasoning_effort to low or medium, because the model thinks at the highest level by default, then read the reply from the first choice.

What Qwen3.8 27B is

Qwen3.8 27B is a dense multimodal model from the Qwen team. It reads text, images and video and writes text. Qwen built it for coding, office work, research and long agent runs. It carries Apache 2.0, while the larger Qwen3.8 flagship stays on a paid API under a custom license.

What it does well

Drives a coding agent through a terminal. It plans steps and reacts to shell output.
Answers questions on a diagram, a scanned document or an hours-long video.
Returns bounding boxes for objects in an image.

What it is not for

Short, cheap replies on default settings. Thinking runs at the highest effort level and spends many reasoning tokens on a plain prompt.
Work that needs an image, audio or video as the answer. The model writes text only.

Calling it

Post to /chat/completions and set model to Qwen/Qwen3.8-27B-FP8. Thinking runs by default at the xhigh effort level, so send reasoning_effort as low or medium for routine work, or set enable_thinking to false to switch it off. Set max_tokens high enough for the reasoning text, which counts against the limit.

Qwen/Qwen3.8-27B-FP8 /chat/completions

SDKs and languages

Qwen3.8 27B answers on an HTTP endpoint, so any language that can post JSON reaches it. These are the ones the box above writes out.

OpenAI SDK: PythonTypeScriptGoJavaC#PHPRust
Anthropic SDK: PythonTypeScriptGoJavaC#PHPRust
HTTP: cURLPythonTypeScriptGoRustJavaC#PHP

Specifications

Author	Qwen
Category	language
Accepts	text, image, video
Returns	text
Total parameters	27.8B
Active parameters	dense, all of them
Score	33.9 on Artificial Analysis intelligence index
Languages	not published
Precision served	FP8
Context	1M tokens
Vector	n/a
License	Apache 2.0

Price

Input

$0.180

per million tokens

Output

$1.980

per million tokens

Against OpenRouter

10%

below the cheapest of 15 providers

The cheapest OpenRouter price for this model was $0.200 in and $2.200 out, with a median of $0.250 and $2.550 across the comparable providers. Read on 2026-09-17.

License

The weights carry Apache 2.0. It permits commercial use, modification and redistribution, with no user cap and no revenue cap, so no extra vendor terms need reading before you ship.

Parameters: 304.2B, 13B
Context: 1.3M
Input per 1M: $0.027
Output per 1M: $0.117
License: MIT