<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>DeepSeek on Learn by Tanhdev</title><link>https://learn.tanhdev.com/tags/deepseek/</link><description>Recent content in DeepSeek on Learn by Tanhdev</description><image><title>Learn by Tanhdev</title><url>https://learn.tanhdev.com/vesviet.png</url><link>https://learn.tanhdev.com/vesviet.png</link></image><generator>Hugo</generator><language>vi</language><lastBuildDate>Sun, 26 Apr 2026 07:30:00 +0700</lastBuildDate><atom:link href="https://learn.tanhdev.com/tags/deepseek/index.xml" rel="self" type="application/rss+xml"/><item><title>Tech Radar 26/04/2026: Sê-ri DeepSeek-V4 Trình Làng — Ngậm 1M Context, Hầu Hạ Agents, kẹp Bãi Efficiency Rặt Open Source</title><link>https://learn.tanhdev.com/radar/2026-04/radar-2026-04-26/</link><pubDate>Sun, 26 Apr 2026 07:30:00 +0700</pubDate><guid>https://learn.tanhdev.com/radar/2026-04/radar-2026-04-26/</guid><description>Tay DeepSeek vỗ bóc tem xuất chuồng sê-ri model DeepSeek-V4, cõng theo lão Pro 1.6T parameter kẹp gã Flash đè bẹp dí rãnh efficiency. Lặn lút moi móc cái context window 1-triệu-token của tụi nó, rổ agentic integrations (đồ nghề cắm agent) nhà trồng, kẹp chuyện bãi legacy endpoints (endpoint đời Tống) bị vứt sọt rác chĩa mũi giáo kiểu gì vô mặt đám developers.</description></item></channel></rss>