<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Moonshot on Vibe Coding</title><link>https://vibecoding.rest/tags/moonshot/</link><description>Recent content in Moonshot on Vibe Coding</description><generator>Hugo</generator><language>en</language><atom:link href="https://vibecoding.rest/tags/moonshot/index.xml" rel="self" type="application/rss+xml"/><item><title>Kimi K2.7 Code</title><link>https://vibecoding.rest/models/kimi-k2-7-code/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://vibecoding.rest/models/kimi-k2-7-code/</guid><description>&lt;p&gt;Kimi K2.7 Code is Moonshot AI&amp;rsquo;s coding-focused release in the Kimi K2 line, built to carry end-to-end programming tasks across long, multi-turn agent sessions rather than one-shot completions.&lt;/p&gt;&#10;&lt;h2 id="why-it-stands-out"&gt;Why it stands out&lt;/h2&gt;&#10;&lt;ul&gt;&#10;&lt;li&gt;&lt;strong&gt;Always-on thinking mode&lt;/strong&gt; that preserves full reasoning content across multi-turn conversations, instead of resetting context between steps.&lt;/li&gt;&#10;&lt;li&gt;&lt;strong&gt;Native multimodal mixture-of-experts architecture&lt;/strong&gt;, accepting text and image input in the same coding session.&lt;/li&gt;&#10;&lt;li&gt;&lt;strong&gt;Frontier-competitive coding benchmarks&lt;/strong&gt; at a fraction of the per-token cost of comparable Western frontier models.&lt;/li&gt;&#10;&lt;li&gt;&lt;strong&gt;Steep cache discount&lt;/strong&gt; (roughly 80% off cache-miss input pricing), which rewards long, iterative agent loops that keep re-reading the same context.&lt;/li&gt;&#10;&lt;/ul&gt;&#10;&lt;h2 id="good-for"&gt;Good for&lt;/h2&gt;&#10;&lt;p&gt;Teams running long, agentic coding sessions who want frontier-adjacent reasoning quality without frontier-model pricing.&lt;/p&gt;</description></item></channel></rss>