Model details

Qwen3.8 2.4T A95B

Qwen · qwen3.8-2.4t-a95b

Qwen3.8-2.4T-A95B is an open-source version of Qwen's latest flagship series, released in August 2026. It adopts a sparse MoE architecture, with a total of 2.4 trillion parameters, activating approximately 95 billion per step, coupled with a hybrid attention mechanism, supporting a 1 million Token context. Core benchmarks: GPQA Diamond 92.6, PaperBench 93.0, OSWorld 86.1, BabyVision 82.0, ranking 4th globally on CodeArena.

能调工具会深度思考
Official price¥12.00/ ¥36.00每百万 Token
中转最低1% of official¥0.1198 / ¥0.3593
19家在卖

The best way to buy Qwen3.8 2.4T A95B

The best deal for each of three ways to buy. Tap a card to see every option of that kind.

Pay-as-you-go API · relay stationsFor coding and tools like Cursor and Claude Code
For using the web app and mobile app, no coding
Official subscriptionsFor a genuine official plan in your own name

Qwen3.8 2.4T A95B

Qwen3.8-2.4T-A95B is an open-source version of Qwen's latest flagship series, released in August 2026. It adopts a sparse MoE architecture, with a total of 2.4 trillion parameters, activating approximately 95 billion per step, coupled with a hybrid attention mechanism, supporting a 1 million Token context. Core benchmarks: GPQA Diamond 92.6, PaperBench 93.0, OSWorld 86.1, BabyVision 82.0, ranking 4th globally on CodeArena.

View intro source
Official price · Input¥12.00
Official price · Output¥36.00
Sellers19 家

The best way to buy Qwen3.8 2.4T A95B

The best deal for each of three ways to buy. Tap a card to see every option of that kind.

Pay-as-you-go API · relay stationsFor coding and tools like Cursor and Claude Code
For using the web app and mobile app, no coding
Official subscriptionsFor a genuine official plan in your own name
No reviews yet — write the first oneSee reviews ↓

模型介绍

Qwen3.8-2.4T-A95B is an open-source version of Qwen's latest flagship series, released in August 2026. It adopts a sparse MoE architecture, with a total of 2.4 trillion parameters, activating approximately 95 billion per step, coupled with a hybrid attention mechanism, supporting a 1 million Token context. Core benchmarks: GPQA Diamond 92.6, PaperBench 93.0, OSWorld 86.1, BabyVision 82.0, ranking 4th globally on CodeArena.

模型基础信息

厂商
通义千问
模型类型
文本与多模态
能接收什么
文本可以发给它的内容类型
能产出什么
文本它能生成的内容类型
能调工具吗
可以能不能自动调用外部函数、查资料、执行动作
怎么计费
按 Token 计费官方按什么单位收钱
现在还能用吗
可以用 · 状态来源
这些资料是哪个地区的资料适用地区
中国大陆厂商专门为中国大陆公布的资料。
它会做什么擅长推理支持结构化输出回答风格可调会写字会调工具

官方价格

输入 · 中国大陆
12 CNY每 100 万 Token
输出 · 中国大陆
36 CNY每 100 万 Token
缓存读取 · 中国大陆
1.5 CNY每 100 万 Token
缓存写入 · 中国大陆
15 CNY每 100 万 Token

中国大陆 / 全球 · 价格附有适用条件

折扣与价格评分按中国大陆的官方价计算

查看价格及适用条件

缓存读取

不分档1.5 CNY每百万 Token
不分档3.34 CNY每百万 Token
缓存方式:显式缓存1 CNY每百万 Token

缓存写入

不分档15 CNY每百万 Token

输入

不分档12 CNY每百万 Token
不分档13.36 CNY每百万 Token

输出

不分档36 CNY每百万 Token
不分档40.08 CNY每百万 Token

中国大陆 / 全球 · 采集于 2026-11-18 · 查看价格来源

它还叫这些名字(15 条)
Qwen3.8-2.4T-A95Bqwen3.8-2.4t-a95b-highqwen3.8-2.4t-a95b-lowqwen3.8-2.4t-a95b-mediumqwen3.8-2.4t-a95b-minimalqwen3.8-2.4t-a95b-non-thinkingqwen3.8-2.4t-a95b-nothinkingqwen3.8-2.4t-a95b-thinkingqwen3.8-2.4t-a95b-thinking-highqwen3.8-2.4t-a95b-thinking-lowqwen3.8-2.4t-a95b-thinking-mediumqwen3.8-2.4t-a95b-thinking-minimalqwen3.8-2.4t-a95b-thinking-xhighqwen3.8-2.4t-a95b-xhigh通义千问-3.8-2.4t-a95b

不同地方写法可能不一样,列在这里方便你对上号。名字太模糊、认不准的不会公开。

User reviews

No reviews yet. If you have used it, share the first honest review and help others avoid pitfalls.

Review this model

Rating
Optional
编辑器加载中…