/en/tags/kimi/Recent content in Kimi on Mi&Bee BlogHugo -- gohugo.ioen蓝宝石的傻话Mon, 23 Mar 2026 00:00:00 +0000/en/posts/aihelper/domestic-llm-cost-comparison/Mon, 23 Mar 2026 00:00:00 +0000/en/posts/aihelper/domestic-llm-cost-comparison/<h2 id="overview">Overview</h2> <p>This article compares the resource requirements and usage costs of three major domestic LLMs, helping developers choose the right solution for their scenarios.</p> <table> <thead> <tr> <th>Model</th> <th>Vendor</th> <th>Architecture</th> <th>Minimum Deployable VRAM</th> <th>API Available</th> </tr> </thead> <tbody> <tr> <td>GLM-5</td> <td>Zhipu AI</td> <td>Dense (multiple versions)</td> <td>24GB (8B)</td> <td>✅</td> </tr> <tr> <td>Kimi K2.5</td> <td>Moonshot AI</td> <td>MoE (undisclosed)</td> <td>24GB (lightweight)</td> <td>✅</td> </tr> <tr> <td>MiniMax M2.7</td> <td>MiniMax</td> <td>MoE 230B</td> <td>Not yet open-sourced</td> <td>✅</td> </tr> </tbody> </table> <h2 id="glm-5-zhipu-ai">GLM-5 (Zhipu AI)</h2> <h3 id="versions--hardware-requirements">Versions &amp; Hardware Requirements</h3> <p>GLM-5 offers 4 parameter versions, making it the widest-coverage domestic LLM currently available.</p>