Qwen3.8 2.4T A95B API

Run Qwen's largest open weight model for coding and long agent runs, from $1.800 per million input tokens.

1M token context
2.4T total parameters
FP8 precision served
Qwen3.8 Max 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/Qwen3.8-2.4T-A95B",
    messages=[
        {
            "role": "user",
            "content": "Why do sheep count people?"
        }
    ],
)

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

How to use the Qwen3.8 2.4T A95B 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/Qwen3.8-2.4T-A95B.
3
Call /chat/completions with your messages and a high max_tokens, and the response returns one choice that holds the reasoning trace and the answer.

What Qwen3.8 2.4T A95B is

Qwen3.8 2.4T A95B is a mixture of experts model from Qwen. It holds 2.4T parameters and runs 95B of them per token. Qwen built it for coding, professional work, research and long agent tasks, and it is the first Max class model the team released with open weights. The hosted Qwen3.8-Max service adds vision input, a non thinking mode and built in tools. These open weights take text only and always think.

What it does well

Writes and repairs code across a repository over many steps.
Runs an agent loop that plans, calls tools and finishes the task in one pass.
Reads a large document set in one prompt up to 1M tokens.

What it is not for

Images, audio and video. The open weights accept text alone.
A short, cheap answer. Thinking cannot be switched off, so every reply pays

Calling it

Post to /chat/completions with the model string Qwen/Qwen3.8-2.4T-A95B and read it from the OpenAI SDK. Set max_tokens high. The model writes a reasoning trace before the answer, so a small limit cuts the reply off before it starts. Use reasoning_effort to pick low, medium or xhigh, and drop to a lower level when a task does not need deep analysis.

Qwen/Qwen3.8-2.4T-A95B /chat/completions

SDKs and languages

Qwen3.8 2.4T A95B 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
Returns	text
Total parameters	2.4T
Active parameters	95B
Score	40 on Artificial Analysis intelligence index
Languages	not published
Precision served	FP8
Context	1M tokens
Vector	n/a
License	Qwen3.8 Max

Price

Input

$1.800

per million tokens

Output

$5.400

per million tokens

Against OpenRouter

10%

below the cheapest of 6 providers

The cheapest OpenRouter price for this model was $2.000 in and $6.000 out, with a median of $2.000 and $6.000 across the comparable providers. Read on 2026-09-17.

License

The Qwen3.8 Max license grants use, modification, distribution, hosting, fine tuning and sale at no charge, under two conditions: a product above the stated user or revenue threshold must display the model name in its interface, and a model as a service or AI work assistant business above the stated revenue threshold must obtain a separate license from Qwen. This is not a standard open source license, so read the license file before you ship.

Parameters: 763.2B, 13B
Context: 1M
Input per 1M: $0.135
Output per 1M: $0.540
License: MIT