Salamandra 2B Instruct
The original Salamandra 2B Instruct adapts BSC's 2B base for prompts and conversation. It is an early instruction-following research release.
Research release
Model and training
BSC reports instruction tuning on 273,000 examples, with particular attention to Catalan, English and Spanish. The card describes a proof of concept without RLHF alignment. The underlying 24-layer architecture and 256,000-token vocabulary remain. Quantized distributions represent the same instructed release.Salamandra 2B Instruct: official model card
Practical use
Practical guidance. This smaller chat model can support experiments where local memory is limited. Evaluate concise instructions, structured answers and language switching separately. Its early post-training means a fluent answer is not evidence of dependable instruction compliance. Use later releases as comparison points when evaluating a new application.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | salamandra-2b-instructsalamandra-2b-instruct-fp8salamandra-2b-instruct-gptq |
| Context window | 8,192 tokens Maximum sequence length documented for these weights; the prompt and generated text share the window. Salamandra 2B Instruct: official model card |
| License | Apache License 2.0.Salamandra 2B Instruct: official model card |
| Model size | 2,253,490,176 parameters.Salamandra 2B Instruct: official model card |
Related models
- Salamandra 2B
Barcelona Supercomputing Center · Salamandra
Two-billion-parameter multilingual base language model pretrained from scratch on European-language text and code.
Launch date not confirmedDownloadable The original Salamandra 7B Instruct is BSC's first conversational adaptation of its 7B multilingual base.
Launch date not confirmedResearch releaseDownloadable