github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 16:40:08 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #12313] [Feature Request] Q6_K for Qwen2.5-VL-32B to fill 32 GB VRAM gap (single-file GGUF preferred) #70239

New Issue
Open
opened 2026-05-04 20:45:31 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-05-04 20:45:31 -05:00
Owner
Copy Link

Originally created by @xl2014 on GitHub (Sep 17, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12313

  1. 32 GB VRAM Utilization Gap

    Q4_K_M ≈ 21 GB → 30 %+ idle, wasted compute
    Q8_0 ≈ 35 GB → exceeds single 32 GB GPU, forces offload or smaller batch
    Q6_K lands at 28–30 GB, perfectly filling 32 GB VRAM while keeping quality

  2. Missing Official Option
    Ollama library only ships q4_K_M and q8_0 tags; no Q6_K middle ground, leaving 32 GB owners stuck between “too empty” and “too large”.

  3. Implementation Request

    Tag name: qwen2.5-vl:32b-q6_k
    Bundle upstream language + vision_encoder into single GGUF (like llava:34b-v1.6-q6_k) for one-command install
    Users with 32 GB GPUs can then ollama pull and immediately run a vision model that uses almost all VRAM—no manual merge needed.

Summary
Adding Q6_K closes the 32 GB VRAM gap in one shot and gives a seamless single-file experience—two birds with one stone.

Originally created by @xl2014 on GitHub (Sep 17, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12313 1. 32 GB VRAM Utilization Gap Q4_K_M ≈ 21 GB → 30 %+ idle, wasted compute Q8_0 ≈ 35 GB → exceeds single 32 GB GPU, forces offload or smaller batch Q6_K lands at 28–30 GB, perfectly filling 32 GB VRAM while keeping quality 2. Missing Official Option Ollama library only ships q4_K_M and q8_0 tags; no Q6_K middle ground, leaving 32 GB owners stuck between “too empty” and “too large”. 3. Implementation Request Tag name: qwen2.5-vl:32b-q6_k Bundle upstream language + vision_encoder into single GGUF (like llava:34b-v1.6-q6_k) for one-command install Users with 32 GB GPUs can then ollama pull and immediately run a vision model that uses almost all VRAM—no manual merge needed. Summary Adding Q6_K closes the 32 GB VRAM gap in one shot and gives a seamless single-file experience—two birds with one stone.
GiteaMirror added the feature request label 2026-05-04 20:45:31 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label feature request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#70239
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?