[GH-ISSUE #9921] Deepseek says it's ChatGPT #68552

Closed
opened 2026-05-04 14:25:43 -05:00 by GiteaMirror · 4 comments
Owner

Originally created by @twotwoiscute on GitHub (Mar 21, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9921

What is the issue?

Believe it or not , Deepseek-v3 says it's ChatGPT, which is definitely not the one when you invoke though the api provided from Deepseek.
Image

Relevant log output


OS

No response

GPU

No response

CPU

No response

Ollama version

0.6.1

Originally created by @twotwoiscute on GitHub (Mar 21, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9921 ### What is the issue? Believe it or not , Deepseek-v3 says it's ChatGPT, which is definitely not the one when you invoke though the api provided from Deepseek. ![Image](https://github.com/user-attachments/assets/1ec5bffd-20ea-43e9-a2ef-48e4c341b2d2) ### Relevant log output ```shell ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version 0.6.1
GiteaMirror added the bug label 2026-05-04 14:25:43 -05:00
Author
Owner

@dpk-it commented on GitHub (Mar 21, 2025):

this is not a bug, it's all because of the data this model was trained on..

<!-- gh-comment-id:2742091720 --> @dpk-it commented on GitHub (Mar 21, 2025): this is not a bug, it's all because of the data this model was trained on..
Author
Owner

@twotwoiscute commented on GitHub (Mar 21, 2025):

this is not a bug, it's all because of the data this model was trained on..

Well, as I said, this is no the one when you invoke though the api provided from Deepseek.

<!-- gh-comment-id:2742104018 --> @twotwoiscute commented on GitHub (Mar 21, 2025): > this is not a bug, it's all because of the data this model was trained on.. Well, as I said, this is no the one when you invoke though the api provided from Deepseek.
Author
Owner

@dpk-it commented on GitHub (Mar 21, 2025):

During training, the model learns to build logical sentences from words. If the dataset that was used for training contained answers generated using ChatGPT, the model learns them and can answer as in your picture. Therefore, the model developers include the model name in the training dataset or use a system prompt so that the model answers this question correctly.

In general, the model may answer this question randomly (if it has not been trained to answer this question with a specific name) and call itself by different names due to how LLM models work.

<!-- gh-comment-id:2742117492 --> @dpk-it commented on GitHub (Mar 21, 2025): During training, the model learns to build logical sentences from words. If the dataset that was used for training contained answers generated using ChatGPT, the model learns them and can answer as in your picture. Therefore, the model developers include the model name in the training dataset or use a system prompt so that the model answers this question correctly. In general, the model may answer this question randomly (if it has not been trained to answer this question with a specific name) and call itself by different names due to how LLM models work.
Author
Owner

@pdevine commented on GitHub (Mar 21, 2025):

As mentioned, this is just the way that model works and not an Ollama problem per se.

<!-- gh-comment-id:2742521518 --> @pdevine commented on GitHub (Mar 21, 2025): As mentioned, this is just the way that model works and not an Ollama problem per se.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#68552