[GH-ISSUE #10994] Ollama Loads Unsloth Qwen3-14B Model with Unexpected High Memory Usage #84934

Closed
opened 2026-05-09 22:06:27 -05:00 by GiteaMirror · 2 comments
Owner

Originally created by @NEWbie0709 on GitHub (Jun 6, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/10994

When using Ollama to load the Unsloth Qwen3-14B-GGUF model (hf.co/unsloth/Qwen3-14B-GGUF:Q4_K_M), the memory usage spikes to 34 GB, while the standard Ollama version of the same model (qwen3:14b, also Q4_K_M) only uses around 10 GB.

Related Issue on Unsloth
I’ve raised this with the Unsloth team here for further insight:
issue

Originally created by @NEWbie0709 on GitHub (Jun 6, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/10994 When using Ollama to load the Unsloth Qwen3-14B-GGUF model (hf.co/unsloth/Qwen3-14B-GGUF:Q4_K_M), the memory usage spikes to 34 GB, while the standard Ollama version of the same model (qwen3:14b, also Q4_K_M) only uses around 10 GB. Related Issue on Unsloth I’ve raised this with the Unsloth team here for further insight: [issue](https://github.com/unslothai/unsloth/issues/2696)
Author
Owner

@rick-github commented on GitHub (Jun 6, 2025):

The unsloth models usually have an embedded num_ctx in the Modelfile. Remove it and the memory use will be reduced.

<!-- gh-comment-id:2948263257 --> @rick-github commented on GitHub (Jun 6, 2025): The unsloth models usually have an embedded `num_ctx` in the Modelfile. Remove it and the memory use will be reduced.
Author
Owner

@NEWbie0709 commented on GitHub (Jun 6, 2025):

@rick-github Thanks, that solved the issue!

<!-- gh-comment-id:2948511165 --> @NEWbie0709 commented on GitHub (Jun 6, 2025): @rick-github Thanks, that solved the issue!
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#84934