Safeguard

Alibaba Cloud (Qwen)

Qwen3.8 2.4T-A95B

qwen/qwen3.8-2.4t-a95b

Request access

Not served yet. Ask for access and we will tell you when it is live.

Large open-source Qwen3.8 model served on Model Studio, with thinking and non-thinking modes.

Price

Input
$2 per 1M tokens
Output
$6 per 1M tokens
Notes
International (Singapore), 0 to 1M input tokens, both modes.
Source
Alibaba Cloud (Qwen) pricing , last verified October 8, 2026

No markup on model usage: you pay the provider's list price.

Model

Context
Not published
Modality
Text
Input
text
Output
text
Weights
Open weights
Tags
open-weights, reasoning

Data policy

Prompts go to
Alibaba Cloud Model Studio (prices shown are for the Singapore / International deployment)
Retention
Alibaba Cloud states it never uses your data for model training, Model Studio stores data generated from model calls as required by law (no period stated)
Zero data retention
Not stated by the provider
Region
Singapore (International) for these prices, other regions such as China (Beijing), Hong Kong, Frankfurt, Virginia and Tokyo have separate price lists

Summarised from the provider's published terms. The provider's own policy is what applies.

Call it with any OpenAI SDK

Available when the model is live

from openai import OpenAI

client = OpenAI(
    base_url="https://api.safeguard.sh/v1",
    api_key="SAFEGUARD_API_KEY",
)

response = client.chat.completions.create(
    model="qwen/qwen3.8-2.4t-a95b",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Base URL https://api.safeguard.sh/v1, model qwen/qwen3.8-2.4t-a95b.

Self-healing security runs on Safeguard.

Your first fix PR is minutes away.

No sales call required, even your agent can complete the purchase over MCP.