公司指南

騰訊混元

Hunyuan,現在常被簡稱為 Hy,是騰訊的基礎模型家族,涵蓋大型語言模型、影片生成與 3D 資產生成。

聆聽此指南AI 語音敘述 · ~8 min

本頁閱讀時間 9 分鐘
  1. 概述
  2. 重點摘要
  3. 深入探討
  4. 像工程師一樣閱讀 Hunyuan 模型卡
  5. 現實世界的實施
  6. 騰訊混元的未來
  7. 付諸實踐
  8. 資料來源與延伸閱讀
  9. 不斷探索
  10. 常見問題

概述

該模型於 2023 年 9 月首次對企業開放,為騰訊的產品(如元寶助手)提供支援。其開放權重的版本(包括 2026 年 8 月發布的 7,700 億參數 Hy4 預覽版)使其成為中國最著名的模型系列之一。

重點摘要

  1. Hunyuan(越來越常被稱為 Hy)是騰訊旗下的語言、影片與 3D 基礎模型系列,於 2023 年 9 月 7 日首次對企業開放。
  2. 於 2024 年 11 月發布的 Hunyuan-Large 是一個混合專家模型(MoE),總參數為 3,890 億,活躍參數為 520 億。
  3. 於 2026 年 4 月 23 日開源的 Hy3 預覽版擁有 2,950 億總參數、210 億活躍參數以及 256,000 個 Token 的上下文視窗。
  4. 於 2026 年 8 月 28 日在 Apache 2.0 授權下發布的 Hy4 預覽版,擁有 7,700 億總參數、490 億活躍參數以及超過一百萬個 Token 的上下文長度。
  5. 大多數已發布的 Hunyuan 效能數據均由騰訊自行報告,尚未經過獨立驗證。

深入探討

Hunyuan is the brand Tencent, the Chinese technology company that owns WeChat, uses for its in-house foundation models. A foundation model is a large neural network trained on broad data that can then be adapted to many tasks. The Hunyuan family is not one model but several lines: general large language models for text and code, image generation models, the HunyuanVideo line for video and the Hunyuan3D line for 3D assets. Recent model cards and announcements use the shorter name Hy, as in Hy3 and Hy4, and describe the developer as the Tencent Hy team. The models serve two audiences: Tencent's own products, such as the Yuanbao chatbot and the CodeBuddy and WorkBuddy assistants, and outside developers through Tencent Cloud and open-weight downloads.

The first public milestone came on 7 September 2023, when Tencent demonstrated Hunyuan in Shenzhen and opened it to enterprise customers. Reuters reported that Tencent described the model as having more than 100 billion parameters, the adjustable values a network learns in training, and as trained on more than 2 trillion tokens, the word fragments a language model reads. Tencent claimed Hunyuan hallucinated, meaning produced confident false statements, 30 percent less than Meta's Llama 2, and said it already underpinned more than 50 internal products; Reuters noted these claims were not independently verified. The launch came in a crowded field: Tencent vice president Jiang Jie said China already had more than 130 large language models by July 2023.

In November 2024 Tencent made its most significant early open release, Hunyuan-Large. Its arXiv paper describes a mixture-of-experts (MoE) model with 389 billion total parameters but only 52 billion activated for each token. In an MoE model, each layer contains many smaller expert sub-networks and a router picks a few of them for each token, so the model can store a lot of knowledge while spending compute like a much smaller model. The paper reports a context window of up to 256,000 tokens, heavy use of synthetic training data and compression of the key-value cache, the memory a transformer keeps about earlier tokens, to make long inputs cheaper. Code and checkpoints were released publicly.

Tencent also built a strong presence in open visual generation. HunyuanVideo, posted in December 2024, is a video generation model with over 13 billion parameters, which its authors described as the largest open-source model of its kind at the time; in professional human evaluations they report it outperformed Runway Gen-3, Luma 1.6 and three leading Chinese video models. Hunyuan3D 2.0, posted in January 2025, splits 3D generation into two models: Hunyuan3D-DiT, a flow-based diffusion transformer that generates a bare shape matching a condition image, and Hunyuan3D-Paint, which generates high-resolution texture maps for that mesh or for one made by hand. A diffusion transformer is an image or shape generator that starts from noise and removes it step by step under the control of a transformer network. The authors report it beats earlier open and closed systems on geometry detail and texture quality, though these are self-reported results.

The language-model line was then rebuilt. Hy3 preview, open-sourced on 23 April 2026, is described on its model card as the first model trained on the team's rebuilt infrastructure. Its model card lists 295 billion total parameters with 21 billion active, 192 experts of which 8 are active per token, a 256,000-token context window and a choice of reasoning effort from none to high. It reports 87.2 on the GPQA Diamond science benchmark and 74.4 percent of tasks resolved on SWE-bench Verified, a test of fixing real software issues. The final Hy3 launched on 6 July 2026, with more post-training compute and better data according to Caixin, which also reported Tencent's claim of a 90 percent completion rate on agent tasks in internal applications.

Hy4 preview followed on 28 August 2026. Tencent's announcement lists 770 billion total parameters, 49 billion active and a context window above one million tokens, with API prices of about 0.83 US dollars per million input tokens and 2.50 dollars per million output tokens. The Hugging Face model card describes 78 layers, 256 routed experts plus one shared expert with the top 8 routed experts used per token, and an Apache 2.0 license, a permissive open-source license that allows commercial use. It reports scores such as 92.3 on GPQA Diamond and 65.7 on SWE-bench Pro. In an internal blind test of 203 engineering tasks judged by 163 experts, Tencent reported an average rating of 2.99 out of 4, against 2.92 for GLM-5.3 and 2.94 for Kimi K3.

Compared with other Chinese model families, including GLM and Kimi, which Tencent itself uses as reference points, Hunyuan stands out less for a single architectural idea than for breadth: Tencent maintains competitive language, video and 3D lines at the same time and ties them to products with very large user bases. Its recent language models follow the same broad recipe as many peers, large sparse MoE transformers with long context and adjustable reasoning, and the Hy4 card names DeepSeek Sparse Attention as part of its design. Licensing has varied by release, which matters for builders: Hy3 preview uses the Tencent Hy Community License, while Hy4 preview uses Apache 2.0.

The main limitations are about verification and access. Most headline numbers, from the 2023 hallucination claim to the Hy4 blind evaluation, come from Tencent itself, and the comparison models are chosen by Tencent. Benchmarks such as GPQA and SWE-bench measure narrow skills and can be affected by training data overlap. Running the largest open models is expensive: Tencent's own deployment examples for Hy3 and Hy4 use eight large-memory GPUs, and 295 billion parameters stored in 16-bit precision alone take roughly 590 GB. For most users, practical access is through Tencent's apps or paid APIs rather than self-hosting.

04工作範例

像工程師一樣閱讀 Hunyuan 模型卡

  1. 在 Hugging Face 上開啟 Hy3‑preview 模型卡,並留意總參數、活躍參數與上下文長度。

  2. 將活躍參數除以總參數以估算每個 token 執行的網路比例。

  3. 將總參數乘以 2 位元組估算 16 位元權重的記憶體需求,未包含任何鍵值快取。

  4. 檢查授權欄位並閱讀其是否允許您預期的商業用途。

  5. 對 Hy4-preview 卡片重複上述步驟,並比較專家數量、上下文窗口與授權。

  6. 列出哪些基準分數來自騰訊自己的評估,哪些(若有)來自獨立排行榜。

它顯示了什麼

學習者以一頁比較結束,顯示 Hy3 preview 每個 token 約執行其 295 億參數的 7% 並且僅權重就需要約 590 GB,並列出每個版本的授權與基準來源。數字可直接對照模型卡檢查。

現實世界的實施

騰訊於 2023 年 9 月表示 Hunyuan 已作為其 50 多項自有產品與服務的基礎,當時將模型開放給外部企業客戶。

HunyuanVideo,於 2024 年 12 月以開源程式碼發布,是一個擁有 130 億以上參數的文字到影片模型,其作者在專業評估中評價高於 Runway Gen-3 與 Luma 1.6。

Hunyuan3D 2.0 以兩階段流程將單張圖像轉換為帶紋理的 3D 網格,先形狀再紋理,程式碼與權重已發布於 GitHub,供遊戲與設計工作使用。

Hy3,2026 年 7 月的語言模型,在發布當天已整合進騰訊的元寶消費者助手,財訊報導同時指出騰訊削減了開發者使用費。

開發者可使用 vLLM 或 SGLang 的開放權重,以 8 路張量並行執行 Hy4 preview,或透過騰訊雲 TokenHub 與 OpenRouter 呼叫。

騰訊混元的未來

Tencent's own statements point to more frequent releases with a preview-first pattern: its August 2026 Hy4 announcement says the team publishes previews, gathers real-world feedback and then ships official releases, and that more Hy4-series models were expected soon. The same announcement emphasizes agent work, such as coding, office documents and game prototyping, and claims Hy4 preview took part in automated optimization of its own training methods and inference system, a claim that has not been independently assessed. Pricing is also moving: Caixin reported that Tencent cut developer fees when the final Hy3 launched in July 2026, and the Hy4 preview announcement offered two weeks of free use in CodeBuddy and WorkBuddy at launch. Open questions include how Hunyuan models perform on independent evaluations rather than Tencent-selected comparisons, whether the licensing shift to Apache 2.0 with Hy4 preview continues for future releases, and how the visual lines, HunyuanVideo and Hunyuan3D, are integrated with the language models as Tencent moves toward multimodal agents.

付諸實踐

為什麼這很重要

  • 供應商策略. 供應商路線圖會影響您的團隊接下來可以建立的功能。
  • 成本與預算. 商業條款和部署選項會影響長期成本和風險。
  • 風險與安全. 公司激勵措施塑造了產品預設、安全態勢和開放性。

留意

  • 發佈公告可能會超過實際生產工作流程的穩定性。
  • API 定價或政策轉變可能會在一夜之間打破假設。
  • 單一供應商依賴性增加了鎖定和遷移成本。

後續步驟

  1. 使用您自己的任務和資料集評估提供者。
  2. 在整合之前查看隱私、安全和法律條款。
  3. 維護跨模型或供應商的後備計劃。
  4. 監控發行說明,以便路線圖的變更不會讓團隊感到意外。

資料來源與延伸閱讀

  1. Reuters via MalaysiaNow中國騰訊推出大型語言 AI 模型,表示對企業開放使用
  2. Hugging Face (Tencent Hy team)騰訊/Hy4-預覽版
  3. arXiv (Tencent)Hunyuan-Large:一個開源 MoE 模型,擁有 520 億個啟用參數,來自騰訊
  4. arXiv (Tencent)HunyuanVideo:大型影片生成模型的系統化框架
  5. arXiv (Tencent)Hunyuan3D 2.0:為高解析度帶紋理 3D 資產生成擴散模型的擴展
  6. Hugging Face (Tencent Hy team)騰訊/Hy3-preview
  7. Caixin Global騰訊推出升級版 Hunyuan 3 AI 模型,提供免費代理功能
  8. Tencent騰訊發布並開源騰訊 Hy4 preview

不斷探索

免費電子報

獲取每日 AI 簡報

每個工作日上午提供三則經過驗證的 AI 新聞,以淺顯易懂的英文撰寫。永久免費,無廣告。

每個工作日一封電子郵件。一鍵取消訂閱。我們絕不出售或分享您的地址。

測試一下自己

參加騰訊 Hunyuan 測驗

每個答案皆有即時回饋,通過課程後可獲得具備可驗證 ID 的可分享證書。

開始測驗

支持免費的 AI 教育。AI Understanding 是一個 501(c)(3) 非營利組織——絕無廣告,絕無付費牆。 捐款

常見問題

什麼是騰訊 Hunyuan?

Hunyuan,現常簡稱為 Hy,是騰訊旗下的基礎模型家族,涵蓋大型語言模型、影片生成及 3D 資產生成。該系列於 2023 年 9 月首次向企業開放,驅動了騰訊產品(如 Yuanbao 助手),其開放權重發布(包括 2026 年 8 月的 7700 億參數 Hy4 預覽版)使其成為中國最顯著的模型家族之一。

Hunyuan3D 设计用来生产什么?

Hunyuan3D 從文字或圖像輸入產生 3D 資源,對於遊戲和設計很有用。

混元AI助理是騰訊的哪款產品?

騰訊元寶是基於混元模型打造的消費者AI助理。

混源-DiT是什麼類型的模型?

騰訊的 Hunyuan-DiT 論文描述了一種用於文字轉圖像生成的多解析度擴散變壓器。

Hunyuan3D 2.0 在紋理合成階段之前會生成什麼?

Hunyuan3D 2.0 論文描述了幾何生成,隨後進行紋理合成。

騰訊採取了哪些方法來擴大混元的採用範圍?

騰訊開源了多種混元模型,包括 3D 和較小的語言變體,以吸引開發者。