llm-benchmark (ollama-benchmark)
LLM Benchmark for Throughput via Ollama (Local LLMs)
Measure how fast your local LLMs really are—with a simple, cross-platform CLI tool that tells you the tokens-per-second truth.
Installation prerequisites
Working Ollama installation.
To create a virtual environment via python3 -m venv
python3 -m venv .venv
## On Linux and macOS
source .venv/bin/activate
## On Windows Powershell or Cmd
.\.venv\Scripts\activate
To create a virtual environment via uv (For uv virtual environments (recommended for Python 3.13))
uv venv .venv --python 3.13
## On Linux and macOS
source .venv/bin/activate
## On Windows Powershell or Cmd
.\.venv\Scripts\activate
Installation Steps
Depending on your python setup either
pip install llm-benchmark
or
pipx install llm-benchmark
or uv
uv pip install llm-benchmark
Usage for general users directly
llm_benchmark run
Installation and Usage in Video format
It's tested on Python 3.10 and above.
ollama installation with the following models installed
7B model can be run on machines with 8GB of RAM
13B model can be run on machines with 16GB of RAM
Usage explaination
On Windows, Linux, and macOS, it will detect memory RAM size to first download required LLM models.
When memory RAM size is greater than or equal to 4GB, but less than 7GB, it will check if gemma:2b exist. The program implicitly pull the model.
ollama pull deepseek-r1:1.5b
ollama pull gemma:2b
ollama pull phi:2.7b
ollama pull phi3:3.8b
When memory RAM size is greater than 7GB, but less than 15GB, it will check if these models exist. The program implicitly pull these models
ollama pull phi3:3.8b
ollama pull gemma2:9b
ollama pull mistral:7b
ollama pull llama3.1:8b
ollama pull deepseek-r1:8b
ollama pull llava:7b
When memory RAM size is greater than 15GB, but less than 31GB, it will check if these models exist. The program implicitly pull these models
ollama pull gemma2:9b
ollama pull mistral:7b
ollama pull phi4:14b
ollama pull deepseek-r1:8b
ollama pull deepseek-r1:14b
ollama pull llava:7b
ollama pull llava:13b
When memory RAM size is greater than 31GB, it will check if these models exist. The program implicitly pull these models
ollama pull phi4:14b
ollama pull deepseek-r1:14b
ollama pull gpt-oss:20b
Python Poetry manually(advanced) installation
https://python-poetry.org/docs/#installing-manually
For developers to develop new features on Windows Powershell or on Ubuntu Linux or macOS
python3 -m venv .venv
. ./.venv/bin/activate
pip install -U pip setuptools
pip install poetry
Usage in Python virtual environment
poetry shell
poetry install
llm_benchmark hello jason
Example #1 send systeminfo and benchmark results to a remote server
llm_benchmark run
Example #2 Do not send systeminfo and benchmark results to a remote server
llm_benchmark run --no-sendinfo
Example #3 Benchmark run on explicitly given the path to the ollama executable (When you built your own developer version of ollama)
llm_benchmark run --ollamabin=~/code/ollama/ollama
Example #4 run custom benchmark models
- Create a custom benchmark file like following yaml format, replace with your own benchmark models, remember to use double quote for your model name
file_name: "custombenchmarkmodels.yml"
version: 2.0.custom
models:
- model: "deepseek-r1:1.5b"
- model: "qwen:0.5b"
- run with the flag and point to the path of custombenchmarkmodels.yml
llm_benchmark run --custombenchmark=path/to/custombenchmarkmodels.yml
Reference
1 maintainer
Metadata
1 maintainer
Release files for llm-benchmark 0.5.4
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| llm_benchmark-0.5.4.tar.gz | 2.1 MB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| llm_benchmark-0.5.4-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 4.3 MB
Release files / llm_benchmark-0.5.4.tar.gz
| Download URL | llm_benchmark-0.5.4.tar.gz |
|---|---|
| Size | 2.1 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
af3c3564a7958ec48cf134a4d9f23da6aeb57bd5695465000e77cb8fcf803afd
|
|
BLAKE2b-256 checksum How to use checksums |
c6265e0fb120d22438a9202c0fb49654e49250fe4e10eaa4930dc90a61343cc8
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 18, 2026.
Transparency logRelease files / llm_benchmark-0.5.4-py3-none-any.whl
| Download URL | llm_benchmark-0.5.4-py3-none-any.whl |
|---|---|
| Size | 2.1 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
18cf7e62a9c0af953641f8d777191ea4544eff5b45ce8e69363e3455135d5e4e
|
|
BLAKE2b-256 checksum How to use checksums |
f358c5a6d6a1b2cea0c0e188c9ee36e47a3acfeddd256f6c1994adda5eb60e0c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 18, 2026.
Transparency log
