DeepSeek Upgrades DeepSeek-V4-Flash-0731 with Major Agentic and Coding Gains
https://www.marktechpost.com/2026/07/31/deepseek-upgrades-deepseek-v4-flash-0731-with-major-agentic-and-coding-gains/📌 【DeepSeek 更新】V4-Flash-0731 正式版發布：透過重新訓練強化 Agent 與程式碼能力
TL;DR：DeepSeek-V4-Flash-0731 透過重新訓練（re-post-training）提升 Agent 與程式碼效能，並推出具備 DSpark 推測解碼能力的正式版本。
DeepSeek 於 2026 年 7 月 31 日正式發布了 DeepSeek-V4-Flash-0731，這是一個取代先前預覽版的正式版本。值得注意的是，這次的效能提升並非源於架構設計的改變，而是透過重新訓練（re-post-training）來實現。
🧩 架構設計與技術細節
根據 DeepSeek 技術報告，V4-Flash 採用 MoE（Mixture-of-Experts）架構，具備 284B 參數，每次 token 僅啟動 13B 參數。
- MoE 結構：每個 MoE 層包含 1 個共享專家與 256 個路由專家，中間維度為 2048，每次 token 會觸發 6 個路由專家。前三層 MoE 使用 hash routing。
- Attention 機制：採用混合注意力機制，結合了壓縮稀疏注意力（Compressed Sparse Attention, CSA）與高度壓縮注意力（Heavily Compressed Attention, HCA）。
- 連接方式：使用流形約束超連接（Manifold-Constrained Hyper-Connections, mHC）取代傳統的殘差連接（residual connections），其擴展因子為 4，並包含 20 次 Sinkhorn-Knopp 迭代。
- 訓練設定：預訓練使用超過 32T tokens，並採用 Muon 優化器。
📊 DSpark 推測解碼帶來生成加速
此版本在 Hugging Face 上的模型權重包含 DSpark 推測解碼模組（與 DeepSeek-V4-Flash-DSpark 結構一致），總參數達到 304B。
- 效能提升：根據 DSpark 論文，在總體吞吐量相同的情況下，V4-Flash 的每用戶生成速度比 MTP-1 基準線快 60–85%。
- 部署方式：使用 vLLM 時，只需透過 flag
--speculative-config '{"method":"dspark","num_speculative_tokens":7,"draft_sample_method":"greedy"}'即可啟用。
💰 極具競爭力的 API 定價與部署建議
DeepSeek 為 V4-Flash 提供了極具吸引力的價格，這對於需要執行大量 Agent 迴圈（agent loops）的開發者非常友善。
- API 定價：
- 輸入（Cache Miss）：$0.14 / 1M tokens
- 輸入（Cache Hit）：$0.0028 / 1M tokens
- 輸出：$0.28 / 1M tokens
- 註：輸出價格僅約 V4-Pro ($0.87) 的三分之一。
- 部署限制：API 提供 2,500 的並發限制（concurrency limit）。
- 自架構（Self-hosting）：權重採用 MIT 授權且無限制。雖然每次僅啟動 13B，但所有專家仍需駐留在記憶體中。
- 單機部署範例：使用 vLLM 可以在單臺 4×GB300 節點上運行。
- 量化版本：Unsloth 的 8-bit 版本需 162 GB，3-bit 版本需 103 GB（需約 110 GB 組合 RAM 與 VRAM）。
💡 開發者實務建議
針對 Agentic（代理式）應用場景，DeepSeek 建議設定 temperature = 1.0 與 top_p = 0.95；其餘場景則建議 top_p = 1.0。此外，API 已原生支援 Responses API 格式並針對 Codex 進行了調整，且 reasoning_effort 參數可設定為 low、high 或 max（在 high 與 max 模式下，輸出 token 最高可達 384K）。
⚠️ 限制
- 此更新僅針對 V4-Flash 版本，V4-Pro API 以及 App 與 Web 端模型並未進行更新。
- 目前模型未提供 Jinja chat template，需使用官方提供的
encoding/資料夾中的函式進行訊息處理。
🔗 來源
- 標題：DeepSeek Upgrades DeepSeek-V4-Flash-0731 with Major Agentic and Coding Gains
- 作者／機構：Asif Razzaq @ MarkTechPost
- 連結：https://www.marktechpost.com/2026/07/31/deepseek-upgrades-deepseek-v4-flash-0731-with-major-agentic-and-coding-gains/
#DeepSeek #V4Flash #LLM #MoE #DSpark #MachineLearning #AI #Coding #AgenticAI #GenerativeAI
原始資料 MarkTechPost · 收集於 2026-08-01
摘要原文
DeepSeek published DeepSeek-V4-Flash-0731 on Hugging Face and moved the official V4-Flash API into public beta on July 31, 2026. The model card is explicit that this is the official release superseding the preview, and that the architecture and size are unchanged. The gains come from re-post-training, not a new design. The checkpoint ships with the DSpark speculative decoding module attached, matching the structure of DeepSeek-V4-Flash-DSpark . Hugging Face reports 304B parameters for the repo, which includes that draft module on top of the 284B base. On the API side, deepseek-v4-flash now natively supports the Responses API format and is adapted for Codex. The V4-Pro API and the app and web models were not updated. Yes, in two very different ways. Via API, it is deployable by almost anyone: DeepSeek’s pricing page lists deepseek-v4-flash at $0.14 per 1M input tokens on a cache miss, $0.0028 on a cache hit, and $0.28 per 1M output tokens, with a 2,500 concurrency limit. That is roughly a third of deepseek-v4-pro output pricing ($0.87). Seed-stage startups, indie developers, and internal platform teams can run agent loops at this price without a GPU budget. Via self-hosting, the bar is much higher : The weights are MIT-licensed and ungated, but every expert stays resident in memory even though only 13B activate per token. DeepSeek’s vLLM example serves it on a single 4×GB300 node. Unsloth’s dynamic GGUFs put the lossless 8-bit build at 162 GB and a 3-bit build at 103 GB, needing roughly 110 GB of combined RAM plus VRAM. Self-hosting suits mid-size and large enterprises with a serving cluster, or one well-specced workstation at aggressive quantization. Per the DeepSeek-V4 technical report , V4-Flash is a 284B-parameter MoE with 13B activated per token and a 1M-token context window. Each MoE layer holds 1 shared expert and 256 routed experts with an intermediate dimension of 2048, and 6 routed experts fire per token. The first three MoE layers use hash routing. Multi-token prediction depth is 1. Attention is hybrid, combining Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA). Manifold-Constrained Hyper-Connections (mHC) replace conventional residual connections, with expansion factor 4 and 20 Sinkhorn-Knopp iterations. Pre-training used more than 32T tokens and the Muon optimizer. The paper’s headline efficiency figure — 27% of single-token inference FLOPs and 10% of KV cache versus DeepSeek-V3.2 at 1M context — is stated for V4-Pro, not Flash. <!– EMBED HERE: paste wordpress-embed.html into a Custom HTML block –> All figures below are DeepSeek-reported, from the 0731 model card. DSpark is enabled with one vLLM flag: --speculative-config '{"method":"dspark","num_speculative_tokens":7,"draft_sample_method":"greedy"}' . The DSpark paper reports 60–85% faster per-user generation on V4-Flash versus the MTP-1 baseline at matched aggregate throughput. There is no Jinja chat template. DeepSeek ships an encoding/ folder with encode_messages and parse_message_from_completion_text instead. reasoning_effort takes low , high , or max . DeepSeek recommends temperature = 1.0 , top_p = 0.95 for agentic use and 1.0 otherwise, with up to 384K output tokens at high and max . Check out the Model Update on HF . 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 DeepSeek Upgrades DeepSeek-V4-Flash-0731 with Major Agentic and Coding Gains appeared first on MarkTechPost .
由 tencent/hy3:free 自動生成