Pharia 1 LLM 7B Control
Pharia 1 Control is a 7B instruction model designed for concise, length-controlled responses. It leaves application-specific alignment to the developer.
Model design
The decoder has 27 layers and a 128,000-entry Unigram tokenizer. Public weights use the Open Aleph License; commercial deployment is separately offered.Pharia 1 language model card
Practical use
Practical guidance. Use Control when the application needs a specific response policy or domain adaptation. Keep language, output length and factual support in the evaluation. Do not assume the absence of default alignment makes a model more accurate or better suited to every task.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable, Provider product |
| Model identifiers | Aleph-Alpha/Pharia-1-LLM-7B-controlAleph-Alpha/Pharia-1-LLM-7B-control-hf |
| Context window | 8,192 tokens Published checkpoint context. Pharia 1 language model card |
| License | Open Aleph License for non-commercial research and education.Pharia 1 language model card |
| Languages | Trained in seven languages; evaluated in English, German, French and Spanish.Pharia 1 language model card |
Related models
- Pharia 1 LLM 7B Control Aligned
Aleph Alpha · Pharia 1
Safety-aligned seven-billion-parameter Pharia 1 model designed for chat applications.
Launch date not confirmedDownloadable · Provider product