[GH-ISSUE #6600] In ollama, do these llama3.1 models refer to the pretrained basic models or instruction tuned models? #4157

Closed
opened 2026-04-12 15:04:43 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @icecream-and-tea on GitHub (Sep 3, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/6600

What is the issue?

the model id is llama3.1:8b that it seems to refer to the pretrained basic model, but in the ollama model library, the performance of llama3.1:8b is consistent with the finetuned model

llama3.1 huggingface performance
image

the ollama library introduction of llama3.1
image

OS

No response

GPU

No response

CPU

No response

Ollama version

No response

Originally created by @icecream-and-tea on GitHub (Sep 3, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/6600 ### What is the issue? the model id is llama3.1:8b that it seems to refer to the pretrained basic model, but in the ollama model library, the performance of llama3.1:8b is consistent with the finetuned model llama3.1 huggingface performance ![image](https://github.com/user-attachments/assets/35485fce-b185-4286-b827-6dd6ac1a6c7d) the ollama library introduction of llama3.1 ![image](https://github.com/user-attachments/assets/e614e393-7d9f-49e7-9174-b7d9b09d45cc) ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version _No response_
GiteaMirror added the question label 2026-04-12 15:04:43 -05:00
Author
Owner

@rick-github commented on GitHub (Sep 3, 2024):

The short form model name in ollama usually refers to the instruction tuned model. The base model normally has the word text or base in the tag, eg llama3.1:8b-instruct-q4_0 vs llama3.1:8b-text-q4_0

<!-- gh-comment-id:2327239018 --> @rick-github commented on GitHub (Sep 3, 2024): The short form model name in ollama usually refers to the instruction tuned model. The base model normally has the word `text` or `base` in the tag, eg [llama3.1:8b-instruct-q4_0](https://ollama.com/library/llama3.1:8b-instruct-q4_0) vs [llama3.1:8b-text-q4_0](https://ollama.com/library/llama3.1:8b-text-q4_0)
Author
Owner

@mxyng commented on GitHub (Sep 3, 2024):

as @rick-github mentioned, pretrained models generally contain text or base in the tag name

<!-- gh-comment-id:2327609214 --> @mxyng commented on GitHub (Sep 3, 2024): as @rick-github mentioned, pretrained models generally contain `text` or `base` in the tag name
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#4157