SpeakLeash/ bielik-11b-v3.1-instruct:Q6_K

100 Downloads Updated yesterday

Polish LLM - Bielik-11B-v3.1-Instruct ~ by SpeakLeash a.k.a Spichlerz!

ollama run SpeakLeash/bielik-11b-v3.1-instruct:Q6_K

curl http://localhost:11434/api/chat \
  -d '{
    "model": "SpeakLeash/bielik-11b-v3.1-instruct:Q6_K",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='SpeakLeash/bielik-11b-v3.1-instruct:Q6_K',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'SpeakLeash/bielik-11b-v3.1-instruct:Q6_K',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated yesterday

yesterday

21d0bcd7027a · 9.2GB

model

archllama

·

parameters11.2B

·

quantizationQ6_K

9.2GB

params

{ "stop": [ "<|im_end|>", "<|im_start|>" ], "temperature": 0.1 }

77B

Readme

Bielik-11B-v3.1-Instruct

Bielik-11B-v3.1-Instruct is an 11-billion-parameter multilingual model developed by SpeakLeash and ACC Cyfronet AGH. It covers 56 languages, with an emphasis on European languages and Polish, and supports conversational use and tool calling.

ollama run SpeakLeash/bielik-11b-v3.1-instruct:Q4_K_M

Original model and full model card · GGUF files on Hugging Face

Available versions

Tag	Approximate download size
`Q4_K_M`	6.7 GB
`Q5_K_M`	7.9 GB
`Q6_K`	9.2 GB
`Q8_0`	11 GB
`bf16`	22 GB

All versions have a 32K context window. For most local setups, start with Q4_K_M; choose a larger version if you have more memory available.

Run from a GGUF file

The GGUF files contain the model’s chat template. When importing one into Ollama, use a Modelfile like this, changing FROM to the path of your file:

FROM ./Bielik-11B-v3.1-Instruct.Q4_K_M.gguf

PARAMETER stop "<|im_end|>"
PARAMETER stop "<|im_start|>"
PARAMETER temperature 0.1

ollama create bielik-v3.1-local -f Modelfile
ollama run bielik-v3.1-local

The template embedded in the GGUF handles chat turns and tool calls; no separate TEMPLATE line is needed in this Modelfile.

Model information

Developed by: SpeakLeash and ACC Cyfronet AGH
Languages: 56, with a focus on European languages and Polish
Model type: causal decoder-only
Based on: Bielik-11B-v3-Base-20250730
License: Apache 2.0

The model was trained using resources provided through PLGrid under grant PLG/2026/019559, on the Athena and Helios supercomputers and within the Gaia AI Factory, funded by the European Union through the EuroHPC Joint Undertaking.

<p align="center">
  <img src="https://huggingface.co/speakleash/Bielik-11B-v3.1-Instruct/resolve/main/speakleash_cyfronet_gaia.png" alt="SpeakLeash, ACC Cyfronet AGH and Gaia AI Factory" height="60">
</p>

# Bielik-11B-v3.1-Instruct

Bielik-11B-v3.1-Instruct is an 11-billion-parameter multilingual model developed by [SpeakLeash](https://speakleash.org/) and [ACC Cyfronet AGH](https://www.cyfronet.pl/en). It covers 56 languages, with an emphasis on European languages and Polish, and supports conversational use and tool calling.

```bash
ollama run SpeakLeash/bielik-11b-v3.1-instruct:Q4_K_M
```

[Original model and full model card](https://huggingface.co/speakleash/Bielik-11B-v3.1-Instruct) · [GGUF files on Hugging Face](https://huggingface.co/speakleash/Bielik-11B-v3.1-Instruct-GGUF)

## Available versions

| Tag | Approximate download size |
| --- | ---: |
| `Q4_K_M` | 6.7 GB |
| `Q5_K_M` | 7.9 GB |
| `Q6_K` | 9.2 GB |
| `Q8_0` | 11 GB |
| `bf16` | 22 GB |

All versions have a 32K context window. For most local setups, start with `Q4_K_M`; choose a larger version if you have more memory available.

## Run from a GGUF file

The GGUF files contain the model's chat template. When importing one into Ollama, use a Modelfile like this, changing `FROM` to the path of your file:

```text
FROM ./Bielik-11B-v3.1-Instruct.Q4_K_M.gguf

PARAMETER stop "<|im_end|>"
PARAMETER stop "<|im_start|>"
PARAMETER temperature 0.1
```

```bash
ollama create bielik-v3.1-local -f Modelfile
ollama run bielik-v3.1-local
```

The template embedded in the GGUF handles chat turns and tool calls; no separate `TEMPLATE` line is needed in this Modelfile.

## Model information

- **Developed by:** [SpeakLeash](https://speakleash.org/) and [ACC Cyfronet AGH](https://www.cyfronet.pl/en)
- **Languages:** 56, with a focus on European languages and Polish
- **Model type:** causal decoder-only
- **Based on:** [Bielik-11B-v3-Base-20250730](https://huggingface.co/speakleash/Bielik-11B-v3-Base-20250730)
- **License:** Apache 2.0

The model was trained using resources provided through [PLGrid](https://plgrid.pl/) under grant PLG/2026/019559, on the Athena and Helios supercomputers and within the [Gaia AI Factory](https://gaia.plgrid.pl/), funded by the European Union through the [EuroHPC Joint Undertaking](https://www.eurohpc-ju.europa.eu/index_en).

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)