The latest articles on Forem by airay (@swb1225). https://forem.com/swb1225 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3918253%2F91c25113-e605-4741-bb0c-faafac722569.jpg https://forem.com/swb1225 en airay Thu, 07 May 2026 15:32:56 +0000 https://forem.com/swb1225/chinese-llms-are-ridiculously-cheap-why-arent-more-developers-using-them-m3p https://forem.com/swb1225/chinese-llms-are-ridiculously-cheap-why-arent-more-developers-using-them-m3p <p>I've been experimenting with Chinese LLMs recently and the pricing caught me off guard:</p> <div class="table-wrapper-paragraph"><table> <thead> <tr> <th>Model</th> <th>Input (per 1M tokens)</th> <th>Output (per 1M tokens)</th> </tr> </thead> <tbody> <tr> <td>GPT-4o</td> <td>$2.50</td> <td>$10.00</td> </tr> <tr> <td>DeepSeek V4</td> <td>$0.30</td> <td>$0.50</td> </tr> <tr> <td>Qwen3.5-Plus</td> <td>$0.11</td> <td>$0.67</td> </tr> <tr> <td>Qwen-Long</td> <td>$0.07</td> <td>$0.28</td> </tr> </tbody> </table></div> <p>That's <strong>8-20x cheaper</strong> than GPT-4o depending on the model.</p> <h2> What's the Catch? </h2> <p>From my experience, the main barriers for non-Chinese developers are:</p> <ol> <li> <strong>Payment</strong> — Most providers only accept Alipay/WeChat Pay</li> <li> <strong>Documentation</strong> — English docs are limited or missing</li> <li> <strong>API compatibility</strong> — Each provider has its own format</li> <li> <strong>Trust</strong> — Concerns about data handling and compliance</li> </ol> <h2> Quality-wise? </h2> <p>I've been running side-by-side tests. For general tasks (chat, summarization, code assistance), DeepSeek and Qwen hold up surprisingly well. They're not beating GPT-4o on everything, but for the price? Hard to ignore.</p> <h2> I'm Curious </h2> <p>For those of you who have tried Chinese LLMs — what was your experience? And for those who haven't, what's holding you back?</p> <p>Would love to hear different perspectives in the comments.</p> ai discuss llm machinelearning