I have searched the existing issues and discussions.
I am using the latest version of Open WebUI.
Installation Method
Docker
Open WebUI Version
v0.6.25
Ollama Version (if applicable)
No response
Operating System
Ubuntu 24.04
Browser (if applicable)
No response
Confirmation
I have read and followed all instructions in README.md.
I am using the latest version of both Open WebUI and Ollama.
I have included the browser console logs.
I have included the Docker container logs.
I have provided every relevant configuration, setting, and environment variable used in my setup.
I have clearly listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc).
I have documented step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation. My steps:
Start with the initial platform/version/OS and dependencies used,
Specify exact install/launch/configure commands,
List URLs visited, user input (incl. example values/emails/passwords if needed),
Describe all options and toggles enabled or changed,
Include any files or environmental changes,
Identify the expected and actual result at each stage,
Ensure any reasonably skilled user can follow and hit the same issue.
Expected Behavior
Running a chat in the llama-cpp-server-UI produces the same token/s result as running a chat in Open-WebUI.
Actual Behavior
I am getting Token Generation of around 23 t/s with llama-cpp-server directly and only around 18 when using Open-WebUI. The prompt processing is roughly the same at around 450 t/s.
GPU utilization is the same when checking with nvidia-smi (only roughly 10% due to VRAM limitations I guess), but CPU utilization with llama-cpp-server is 100% on the used cores while Open-WebUI only uses 100% on one core and around 85% on the other cores (see Screenshots attached).
Open-WebUI: v0.6.25 running in docker with :main image
llama-cpp: b6297, built from source with -DGGML_CUDA=ON
CUDA toolkit 12.8.93
llama-cpp-server command (With running llama-cpp-server directly, using 6 cores spread on both CCDs was the sweet-spot for token generation speed):
./llama.cpp/llama-server
-hf unsloth/gpt-oss-120b-GGUF:Q4_K_M
-b 4096
-ub 2048
-t 6
-tb 12
-C 0x555
-Cb 0xFFF
--n-cpu-moe 35
--n-gpu-layers 999
-c 0
-fa
--jinja
--reasoning-format none
--temp 1.0
--min-p 0.0
--top-p 1.0
--top-k 0.0
--chat-template-kwargs '{"reasoning_effort":"high"}'
-np 2
-cb
No modifications made to Open-WebUI config.
Logs & Screenshots
CPU-utilization with Open-WebUI calling llama-cpp-server:
CPU-utilization with the same prompt in the llama-cpp-server-UI:
Additional Information
No response
Originally created by @Te-eMster on GitHub (Aug 27, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16959
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
- [x] I am using the latest version of Open WebUI.
### Installation Method
Docker
### Open WebUI Version
v0.6.25
### Ollama Version (if applicable)
_No response_
### Operating System
Ubuntu 24.04
### Browser (if applicable)
_No response_
### Confirmation
- [x] I have read and followed all instructions in `README.md`.
- [x] I am using the latest version of **both** Open WebUI and Ollama.
- [x] I have included the browser console logs.
- [x] I have included the Docker container logs.
- [x] I have **provided every relevant configuration, setting, and environment variable used in my setup.**
- [x] I have clearly **listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup** (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc).
- [x] I have documented **step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation**. My steps:
- Start with the initial platform/version/OS and dependencies used,
- Specify exact install/launch/configure commands,
- List URLs visited, user input (incl. example values/emails/passwords if needed),
- Describe all options and toggles enabled or changed,
- Include any files or environmental changes,
- Identify the expected and actual result at each stage,
- Ensure any reasonably skilled user can follow and hit the same issue.
### Expected Behavior
Running a chat in the llama-cpp-server-UI produces the same token/s result as running a chat in Open-WebUI.
### Actual Behavior
I am getting Token Generation of around 23 t/s with llama-cpp-server directly and only around 18 when using Open-WebUI. The prompt processing is roughly the same at around 450 t/s.
GPU utilization is the same when checking with nvidia-smi (only roughly 10% due to VRAM limitations I guess), but CPU utilization with llama-cpp-server is 100% on the used cores while Open-WebUI only uses 100% on one core and around 85% on the other cores (see Screenshots attached).
### Steps to Reproduce
Hardware: Ryzen 9 7900, 128GB DDR5@4000, RTX3080 10GB
Open-WebUI: v0.6.25 running in docker with :main image
llama-cpp: b6297, built from source with -DGGML_CUDA=ON
CUDA toolkit 12.8.93
llama-cpp-server command (With running llama-cpp-server directly, using 6 cores spread on both CCDs was the sweet-spot for token generation speed):
./llama.cpp/llama-server \
-hf unsloth/gpt-oss-120b-GGUF:Q4_K_M \
-b 4096 \
-ub 2048 \
-t 6 \
-tb 12 \
-C 0x555 \
-Cb 0xFFF \
--n-cpu-moe 35 \
--n-gpu-layers 999 \
-c 0 \
-fa \
--jinja \
--reasoning-format none \
--temp 1.0 \
--min-p 0.0 \
--top-p 1.0 \
--top-k 0.0 \
--chat-template-kwargs '{"reasoning_effort":"high"}' \
-np 2 \
-cb
No modifications made to Open-WebUI config.
### Logs & Screenshots
CPU-utilization with Open-WebUI calling llama-cpp-server:
<img width="321" height="103" alt="Image" src="https://github.com/user-attachments/assets/ec8b9b66-30e0-43b6-82aa-b9de46f27d20" />
CPU-utilization with the same prompt in the llama-cpp-server-UI:
<img width="321" height="103" alt="Image" src="https://github.com/user-attachments/assets/b62093d6-43bc-4414-a8f0-5309b4168ccc" />
### Additional Information
_No response_
GiteaMirror
added the bug label 2026-04-20 00:08:14 -05:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @Te-eMster on GitHub (Aug 27, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16959
Check Existing Issues
Installation Method
Docker
Open WebUI Version
v0.6.25
Ollama Version (if applicable)
No response
Operating System
Ubuntu 24.04
Browser (if applicable)
No response
Confirmation
README.md.Expected Behavior
Running a chat in the llama-cpp-server-UI produces the same token/s result as running a chat in Open-WebUI.
Actual Behavior
I am getting Token Generation of around 23 t/s with llama-cpp-server directly and only around 18 when using Open-WebUI. The prompt processing is roughly the same at around 450 t/s.
GPU utilization is the same when checking with nvidia-smi (only roughly 10% due to VRAM limitations I guess), but CPU utilization with llama-cpp-server is 100% on the used cores while Open-WebUI only uses 100% on one core and around 85% on the other cores (see Screenshots attached).
Steps to Reproduce
Hardware: Ryzen 9 7900, 128GB DDR5@4000, RTX3080 10GB
Open-WebUI: v0.6.25 running in docker with :main image
llama-cpp: b6297, built from source with -DGGML_CUDA=ON
CUDA toolkit 12.8.93
llama-cpp-server command (With running llama-cpp-server directly, using 6 cores spread on both CCDs was the sweet-spot for token generation speed):
./llama.cpp/llama-server
-hf unsloth/gpt-oss-120b-GGUF:Q4_K_M
-b 4096
-ub 2048
-t 6
-tb 12
-C 0x555
-Cb 0xFFF
--n-cpu-moe 35
--n-gpu-layers 999
-c 0
-fa
--jinja
--reasoning-format none
--temp 1.0
--min-p 0.0
--top-p 1.0
--top-k 0.0
--chat-template-kwargs '{"reasoning_effort":"high"}'
-np 2
-cb
No modifications made to Open-WebUI config.
Logs & Screenshots
CPU-utilization with Open-WebUI calling llama-cpp-server:
CPU-utilization with the same prompt in the llama-cpp-server-UI:
Additional Information
No response