Doubao allows computers with independent graphics cards to deploy local large models. The corpus is the file in the computer.

Written by

Audrey Miles

Updated on:June-30th-2025

The courseware teacher was on a business trip today. My roommate said that he went to bed very late. He played with his phone for a while while lying on the bed. After a while, he started snoring loudly. I was really tired today, but I still insisted on finishing the official account before taking a rest. I turned on the computer and started looking for writing materials. After the content was written, I asked DS to generate the cover prompt. I opened Doubao and found that it had updated another function today.

Local File Q&A

The system displays a line of red text: Recommended system configuration: RTX GPU with 8GB or more video memory, i5 processor, 16G memory. The current system configuration is too low to use this function.

Unfortunately, the courseware teacher was on a business trip and only had a laptop with him, which only had an integrated graphics card and could not meet the requirements.

The system also prompts: After turning it on, you can use local models for AI question and answer for local files. The files and conversation content will not be uploaded to the server, and the function supports offline use.

This ensures the security of the data. From the perspective of the operation sequence, it should be the same principle as the previous local deployment of DeepSeek, but it is much simpler to operate. In the past, there was a technical threshold to deploy a local model, but now it is as simple as transferring a file. The system prompts that the local model size is 7G. I don’t know which one is better, it or the DS model. I look forward to your answers after testing.

Although the courseware teachers could not experience it firsthand, they were more eager to get a desktop computer to experience it. If DeepSeek lowered the price of generative artificial intelligence, then Doubao's operation lowered the technical threshold for deploying large local models.

Finally, you can also check out that the number of large models that have passed the review has increased again, and now has exceeded 600 large models that have passed the review.