> ## Documentation Index
> Fetch the complete documentation index at: https://docs.modelstudio.console.alibabacloud.com/llms.txt
> Use this file to discover all available pages before exploring further.

# qwen3.8-2.4t-a95b

> Qwen3.8-2.4T-A95B is the open-source release of Qwen's latest flagship, launched August 2026. Its sparse MoE architecture holds 2.4T total parameters with ~95B activated per step, paired with hybrid attention and a 1M token context window. Key benchmarks: GPQA Diamond 92.6, PaperBench 93.0, OSWorld 86.1, BabyVision 82.0. Ranked 4th on CodeArena.

## Inference Service Provider <span id="h-9f4938f724" /> <span id="sec-d0bb5be87e" />

The inference service provider for `qwen3.8-2.4t-a95b` is Alibaba Cloud Model Studio.

## Model Capabilities <span id="h-158bc80c73" /> <span id="sec-59aa0ca7b4" />

<table><thead><tr><th>Capability</th><th>Support</th><th>Capability</th><th>Support</th></tr></thead><tbody><tr><td><p>Input Modality</p></td><td><p><strong>Text</strong></p></td><td><p>Output Modality</p></td><td><p><strong>Text</strong></p></td></tr><tr><td><p>Model Experience</p></td><td><p>Supported</p></td><td><p>Function Calling</p></td><td><p>Supported</p></td></tr><tr><td><p>Structured Outputs</p></td><td><p>Supported</p></td><td><p>Web Search</p></td><td><p>Supported</p></td></tr><tr><td><p>Prefix Completion</p></td><td><p>Supported</p></td><td><p>Context Caching</p></td><td><p>Supported</p></td></tr><tr><td><p>Batch Inference</p></td><td><p>Unsupported</p></td><td><p>Fine-tuning</p></td><td><p>Unsupported</p></td></tr></tbody></table>

## Context Limits <span id="h-5c93024d31" /> <span id="sec-cbace9322c" />

<table style={{ display: "table", tableLayout: "fixed", width: "100%" }}><colgroup><col style={{ width: "25%" }} /><col style={{ width: "25%" }} /><col style={{ width: "25%" }} /><col style={{ width: "25%" }} /></colgroup><thead><tr><th><p>Parameter</p></th><th><p>Value</p></th><th><p>Parameter</p></th><th><p>Value</p></th></tr></thead><tbody><tr><td><p>Max Input Length</p></td><td><p>991808</p></td><td><p>Max Output Length</p></td><td><p>131072</p></td></tr><tr><td><p>Context Length</p></td><td><p>1000000</p></td><td><p>Max Input Length (Thinking Mode)</p></td><td><p>983616</p></td></tr><tr><td><p>Max Output Length (Thinking Mode)</p></td><td><p>131072</p></td><td><p>Max Chain-of-Thought Length</p></td><td><p>131072</p></td></tr></tbody></table>

## Pricing <span id="h-70ebc1fc1e" /> <span id="sec-fe6d9fe6a5" />

This page only shows the original pricing for model API calls, excluding any limited-time promotions. Visit [Model Studio Console](https://modelstudio.console.alibabacloud.com/ap-southeast-1/model/market) for promotional offers.

<Tabs>
  <Tab title="China (Beijing)">
    <table><thead><tr><th>Billing Item</th><th>Price (USD)</th><th>Unit</th></tr></thead><tbody><tr><td><p>Input</p></td><td><p>1.65</p></td><td><p>Per 1M tokens</p></td></tr><tr><td><p>Output</p></td><td><p>4.951</p></td><td><p>Per 1M tokens</p></td></tr><tr><td><p>Input (Cache Hit)</p></td><td><p>0.206</p></td><td><p>Per 1M tokens</p></td></tr><tr><td><p>Explicit Cache Creation</p></td><td><p>2.063</p></td><td><p>Per 1M tokens</p></td></tr><tr><td><p>Explicit Cache Hit</p></td><td><p>0.137</p></td><td><p>Per 1M tokens</p></td></tr></tbody></table>
  </Tab>

  <Tab title="Singapore">
    Scope: International

    <table><thead><tr><th>Billing Item</th><th>Price (USD)</th><th>Unit</th></tr></thead><tbody><tr><td><p>Input</p></td><td><p>2</p></td><td><p>Per 1M tokens</p></td></tr><tr><td><p>Output</p></td><td><p>6</p></td><td><p>Per 1M tokens</p></td></tr><tr><td><p>Input (Cache Hit)</p></td><td><p>0.25</p></td><td><p>Per 1M tokens</p></td></tr><tr><td><p>Explicit Cache Creation</p></td><td><p>2.5</p></td><td><p>Per 1M tokens</p></td></tr><tr><td><p>Explicit Cache Hit</p></td><td><p>0.17</p></td><td><p>Per 1M tokens</p></td></tr></tbody></table>
  </Tab>
</Tabs>

## Rate Limits <span id="h-bff1f1239f" /> <span id="sec-92c10bdf07" />

<Tabs>
  <Tab title="China (Beijing)">
    This region uses dynamic rate limiting. The TPM limit is tiered by your monthly Model Studio spend, and the RPM limit is high enough that normal use does not trigger throttling. For the tier values, see [Dynamic rate limiting](/en/model-studio/quota-management).
  </Tab>

  <Tab title="Singapore">
    Scope: International

    This region uses dynamic rate limiting. The TPM limit is tiered by your monthly Model Studio spend, and the RPM limit is high enough that normal use does not trigger throttling. For the tier values, see [Dynamic rate limiting](/en/model-studio/quota-management).
  </Tab>
</Tabs>