Open PL LLM Leaderboard

Open PL LLM Leaderboard

A Gradio leaderboard that ranks language models on a set of evaluation datasets, with its name and datasets pointing to Polish. For teams comparing models or running a leaderboard of their own.

Made by speakleash on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space is a leaderboard for large language models, built from the standard Hugging Face leaderboard template. Its name and the datasets it lists point to evaluation in Polish. It tracks models such as the Azurro APT3 1B pair, Starling LM 7B alpha, two zephyr 7B builds and internlm2 1.8B.

Who it suits: Researchers and model builders who want to compare language models on Polish tasks, and teams who want a working base for a leaderboard of their own.

Industry | Who uses it | How they use it | Example
AI research | Language model researchers | Compare models on Polish evaluation sets | Look up how two named models score
Model development | Model builders | Check a new model against others | Submit a model for evaluation through the Space
NLP engineering | Applied NLP teams | Pick a model for Polish text work | Read the scores before choosing a model
Open source | Leaderboard maintainers | Fork the template for another language or task | Change the settings in src and swap the datasets

If you want one model's answers rather than a ranking, a model demo page suits you better.

Where to use it: It runs in two places.

• A web browser, on the Space's page at Hugging Face with nothing to install
• A local copy on Windows, macOS or Linux, which needs Python 3.12

Use case | Who | What they get back
Compare models | Researchers | Scores for each listed model on each evaluation set
Queue a model | Model builders | An evaluation request handled by the backend
Fork the leaderboard | Maintainers | A starting point for a leaderboard of their own

How to set it up: Using it online needs nothing. A local copy needs a few things.

• Python 3.12 and pip
• The packages listed in requirements.txt

1. Install the Hugging Face tools.
2. Download the Space into a folder.
3. Move into that folder.
4. Install Gradio and the packages from requirements.txt.
5. Start the app.

pip install -U huggingface_hub
hf download speakleash/open_pl_llm_leaderboard --repo-type space --local-dir open_pl_llm_leaderboard
cd open_pl_llm_leaderboard
pip install gradio==4.4.0
pip install -r requirements.txt
python app.py

How to use it: Open the Space page and read the ranking table. To adapt it as your own leaderboard, change the variables in src/env and the text in src/about. Result files follow this shape, with one entry per task and metric.

{
    "config": {
        "model_dtype": "torch.float16", # or torch.bfloat16 or 8bit or 4bit
        "model_name": "path of the model on the hub: org/model",
        "model_sha": "revision on the hub",
    },
    "results": {
        "task_name": {
            "metric_name": score,
        },
        "task_name2": {
            "metric_name": score,
        }
    }
}

Example: You want to know how an open 7B model compares with a 1B model on the Polish tasks. You open the Space, find both in the table and read their scores side by side.

For newcomers and experts: A leaderboard runs the same tests on many models and lists the scores. Begin with the table and look at one task at a time.

The backend queues requests and creates request files by itself. If the Space misbehaves, a restart clears the folders it created for the queue and results. To plug in a different evaluation, edit the backend's evaluation step. At the moment that step starts an evaluation suite from EleutherAI.

A model is submitted or listed
The backend runs the evaluation suite
Results are written as files in the given shape
The table shows the scores

Good to know: The README is mostly template text, so the exact tasks and scoring are best read from the page.

• License: Apache 2.0
• Cost to run: none named, and the Space runs on upgraded CPU hardware at Hugging Face
• Data it reaches: the models on the Hub and the evaluation datasets it lists, from the KLEJ and PES sets

On Hugging Face, 13 November 2026

• Likes: 84
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/speakleash/open_pl_llm_leaderboard
• Files: https://huggingface.co/spaces/speakleash/open_pl_llm_leaderboard/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.