Alibaba Qwen Releases Qwen3.8-Max: A 2.4 Trillion Parameter MoE Model and the Most Capable One in the Qwen Family to Date
https://www.marktechpost.com/2026/08/03/alibaba-qwen-releases-qwen3-8-max/📌 【Alibaba Qwen】2.4 兆參數 MoE 模型 Qwen3.8-Max 登場，多模態與 Agent 能力大幅躍升
TL;DR：Qwen3.8-Max 採用 2.4T MoE 架構，提供 1M context window 與強大 Agent 能力，API 已開放。
Alibaba Qwen 團隊正式推出 Qwen3.8-Max，這是目前 Qwen 系列中能力最強大的模型。該模型不僅規模達到 2.4 兆參數，更在多模態（Multimodal）與代理任務（Agentic）表現上展現出顯著的世代進步。
🧩 2.4 兆參數的 MoE 架構與多模態輸入
Qwen3.8-Max 採用 Mixture-of-Experts (MoE) 架構，總參數規模達 2.4 兆（2.4T）。
- 輸入與輸出：支援文本、圖像與影片輸入，並輸出文本。
- 上下文長度：提供 1M token 的上下文視窗。其中最大輸入為 991K tokens；若啟用「思考模式（thinking mode）」，最大輸入會降至 983K tokens。最大輸出則維持在 131K tokens，最大推理預算（reasoning budget）為 262K tokens。
- 部署說明：由於參數規模巨大，2.4T 的權重（weights）屬於多節點資料中心級別的產物。目前已透過 Hosted API 開放使用，且相容於 OpenAI 與 DashScope 介面，開發者只需更改 Base-URL 與 Model-ID 即可快速整合。
📊 基準測試：多模態與 Agent 表現是最大亮點
根據 Alibaba 發布的測試數據，Qwen3.8-Max 在多模態與代理任務上的增長最為顯著。
軟體工程與推理能力：
- Terminal-Bench 2.1：Qwen3.8-Max 取得 86.6 分，領先 Claude Opus 4.8 (84.6)，僅次於 GPT-5.6 Sol (max) (88.8)。
- SWE-bench Pro：取得 67.7 分（對比 Fable 5 的 80.0）。
- FrontierSWE：取得 73.5 分（對比 Fable 5 的 88.8）。
- DeepSWE 1.1：從前代 Qwen3.7-Max 的 21.6 分大幅躍升至 56.6 分。
多模態能力（對標 Qwen3.7- Plus）：
- OSWorld-Verified：86.1
- Parametric CAD Bench：91.5
- OmniDocBench 1.5：92.1
⚠️ 技術限制與觀察 作者指出，多模態測試是與 Qwen3.7- Plus 進行對比而非 Max 版本，這可能會放大世代間的差異。此外，Alibaba 的 RL scaling curve 在接近 4,000 個訓練環境時達到 0.725 的峰值後，隨後會下降至 0.719 與 0.689。
💰 API 價格與功能特性
Qwen3.8-Max 提供完整的開發者工具與功能集。
- 定價結構：
- 輸入：$2.00 / 1M tokens
- 輸出：$6.00 / 1M tokens
- 隱含快取讀取 (Implicit cache reads)：$0.25 / 1M tokens
- 顯式快取建立 (Explicit cache creation)：$2.50 / 1M tokens
- 顯式快取讀取 (Explicit cache reads)：$0.17 / 1M tokens
- 註：使用快取輸入的成本僅為原始輸入的 1/8，因此前綴穩定性（Prefix stability）對成本的影響大於 Prompt 長度。
- 內建工具：Responses API 提供 code_interpreter、web_search、web_extractor、t2i_search 與 i2i_search 五種工具。
- 支援功能：Function calling、Structured outputs、Batches、Prefix completion 與 Fine-tuning。
🎯 實務應用與工程啟示
Qwen3.8-Max 的能力範疇涵蓋了四大產業需求，特別適合需要高複雜度邏輯與多模態理解的場景：
- 軟體工程：用於大規模程式碼庫規模的 Coding Agents。
- 法律與金融：處理長文件審閱與知識庫。
- 媒體與電商營運：進行長影片索引與結構化資料擷取。
- 設計領域：支援多步驟的研究助手。
對於資源有限的開發者，Alibaba 也將推出 Qwen3.8-27B 的開源權重版本，該版本更適合在一般的地端 GPU 硬體上運行。
🔗 來源
- 標題：Alibaba Qwen Releases Qwen3.8-Max: A 2.4 Trillion Parameter MoE Model and the Most Capable One in the Qwen Family to Date
- 作者／機構：Asif Razzaq @ MarkTechPost
- 連結：https://www.marktechpost.com/2026/08/03/alibaba-qwen-releases-qwen3-8-max/
#Alibaba #Qwen #MoE #LLM #Multimodal #MachineLearning #AI #SoftwareEngineering #AgenticAI #LargeLanguageModels
原始資料 MarkTechPost · 收集於 2026-08-03
摘要原文
Alibaba’s Qwen team has made Qwen3.8-Max broadly available and confirmed that its open weights ship next week. A second checkpoint, Qwen3.8-27B, is also going open-weights . Qwen3.8-Max is a 2.4-trillion-parameter mixture-of-experts model. It accepts text, image and video as input and returns text. Yes, but the deployable surface depends on which artifact you are applying. The hosted API is deployable today by any company size. It is OpenAI- and DashScope-compatible , so integration is a base-URL and model-ID change. The open weights are a different matter. At 2.4T total parameters, the checkpoint is a multi-node datacenter artifact. Alibaba has not disclosed the activated-parameter count. Serving cost therefore cannot yet be modeled. Qwen3.8-27B is the checkpoint that fits ordinary on-premise GPU hardware. The published feature set maps cleanly onto four industries. Those are software engineering, legal and financial document review, media and e-commerce operations, and design. Applications include repository-scale coding agents and long-document knowledge bases. Long-video indexing, structured data extraction and multi-step research assistants also fit. The model page lists a 1M-token context window. Maximum input is 991K tokens, dropping to 983K when thinking is enabled. Maximum output is 131K tokens in both modes, and the maximum reasoning budget is 262K tokens. Rate limits are 2M tokens per minute and 15K requests per minute. Pricing is $2.00 per 1M input tokens and $6.00 per 1M output tokens. Implicit cache reads cost $0.25 per 1M tokens. Explicit cache creation is $2.50 and explicit cache reads are $0.17 per 1M tokens. Cached input is eight times cheaper than fresh input. Prefix stability therefore drives cost more than prompt length does. Supported capabilities include function calling , structured outputs , batches, prefix completion and fine-tuning. Five built-in tools ship on the Responses API: code_interpreter , web_search , web_extractor , t2i_search and i2i_search . Alibaba published a full benchmark table with this release. Qwen3.8-Max scores 86.6 on Terminal-Bench 2.1, ahead of Claude Opus 4.8 and Claude Fable 5 at 84.6, behind GPT-5.6 Sol (max) at 88.8. It reports 67.7 on SWE-bench Pro against Fable 5’s 80.0, and 73.5 on FrontierSWE against Fable 5’s 88.8. It leads PaperBench at 93.0 and IFBench at 82.8. GPQA Diamond lands at 92.6, up marginally from Qwen3.7-Max’s 92.4. The clearest gains are multimodal and agentic, not reasoning. It tops most vision rows, including OSWorld-Verified 86.1, Parametric CAD Bench 91.5, and OmniDocBench 1.5 at 92.1. Against its own predecessor the jump is large: DeepSWE 1.1 moves from 21.6 to 56.6, FrontierSWE from 40.7 to 73.5, JobBench from 31.3 to 53.4. Two caveats belong in any honest read. The multimodal table benchmarks against Qwen3.7- Plus , not Qwen3.7-Max, which flatters the generational delta. And Alibaba’s own RL scaling curve peaks at 0.725 near 4,000 training environments, then declines to 0.719 and 0.689. Check out the Technical details , API and Qwen Studio . Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter . Wait! are you on telegram? now you can join us on telegram as well. Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us The post Alibaba Qwen Releases Qwen3.8-Max: A 2.4 Trillion Parameter MoE Model and the Most Capable One in the Qwen Family to Date appeared first on MarkTechPost .
由 tencent/hy3:free 自動生成