SpeakLeashLanguage modelAvailable

Bielik 11B v3.1 Instruct

Summary

Bielik 11B v3.1 Instruct is an open-weight text model with 11 billion parameters. SpeakLeash and ACC Cyfronet AGH list 56 languages, including the EU’s official languages, Arabic, and East Asian languages.

The Apache 2.0 weights are hosted on Hugging Face. The platform requires agreeing to share account information before downloading. The model supports tool calling and reasoning; its card does not state a native maximum context length.

Similar models

More models of the same kind as Bielik 11B v3.1 Instruct, chosen by model line, developer, size, price and release date.

All specifications

Input and output

Input: Text.; Source: Input
Output: Text.; Source: Output
Languages: 56 languages, with an emphasis on European languages and Polish.; Source: Languages
Reasoning: Off by default according to the model card. Supported levels: none, low, medium, high, and max.; Source: Reasoning

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking: Unknown; no source
Watermarking technique: Not known; no source

Licence and access

License: Apache 2.0.; Source: License
Access: Weights on Hugging Face. Access to the files requires agreeing to share account information with Hugging Face. The model is not hosted by Hugging Face Inference Providers.; Source: Access
Shutdown: None announced; no source

Further details

Checkpoint: speakleash/Bielik-11B-v3.1-Instruct; Source: Checkpoint
Architecture and size: Generative text model with 11 billion parameters and a causal decoder architecture.; Source: Architecture and size
Tool calling: Supports tool calls and structured output. For parsed tool-call and reasoning fields, the model card documents parser plugins for vLLM.; Source: Tool calling
Bielik-One: SpeakLeash announces four separate decision models for classification and evaluation, planned for November 12, 2026. The announcement does not give individual model IDs.; Source: Bielik-One
Local serving example: The vLLM example sets --max-model-len 32768. This is a configured runtime value and does not establish the model’s native maximum context window.; Source: Local serving example

Frequently asked questions about Bielik 11B v3.1 Instruct

Can I run Bielik 11B v3.1 Instruct locally?

Yes, the weights are open.

Stated by SpeakLeash

Open weights: yes
License: Apache 2.0.
Parameters: 11B
active per token: n/a

Sources and data date

Type	Evidence and data date
Input	Documented modalities. Retrieved November 11, 2026. Source: SpeakLeash Bielik-11B-v3.1-Instruct model card · SpeakLeash Bielik-11B-v3.1-Instruct model card (retrieved November 11, 2026) · SpeakLeash Bielik-11B-v3.1-Instruct model card · Editorial description reviewed November 11, 2026
Output	Documented modalities. Retrieved November 11, 2026. Source: SpeakLeash Bielik-11B-v3.1-Instruct model card · SpeakLeash Bielik-11B-v3.1-Instruct model card (retrieved November 11, 2026) · SpeakLeash Bielik-11B-v3.1-Instruct model card · Editorial description reviewed November 11, 2026
Additional source	Bielik 11B v3.1 launch announcement (retrieved November 11, 2026) · Bielik 11B v3.1 and DFlash announcement · Editorial description reviewed November 11, 2026