SalamandraTA 2B
SalamandraTA 2B is a small translation model derived from BSC's Salamandra 2B base.
Model and training
BSC continued pretraining on 424 billion parallel-data tokens before translation instruction tuning. The internal translation base is not a separate public release here. This version covers 35 European languages and three language varieties. Its tasks include translation, post-editing, grammar correction and translation evaluation. The card distinguishes its older coverage from later 7B updates.SalamandraTA 2B: official model card
Practical use
Practical guidance. Evaluate complete language pairs, rather than treating a language list as evidence of equal quality in every direction. This model may suit focused translation workflows where local memory matters. Use the translation prompt format and inspect names, numbers and omissions. General conversational ability is not the model's stated purpose.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | salamandraTA-2b-instructsalamandraTA-2B-instruct-GGUF |
| Context window | 8,192 tokens Maximum sequence length documented for these weights; the prompt and generated text share the window. SalamandraTA 2B: official model card |
| License | Apache License 2.0.SalamandraTA 2B: official model card |
| Model size | 2,253,490,176 parameters.SalamandraTA 2B: official model card |
Related models
- SalamandraTA 2B Academic
Barcelona Supercomputing Center · SalamandraTA
SalamandraTA 2B Academic adapts BSC's translation model to academic text.
Launch date not confirmedDownloadable - SalamandraTA 7B v1
Barcelona Supercomputing Center · SalamandraTA
SalamandraTA 7B v1 is the first versioned 7B translation release based on BSC's multilingual Salamandra model.
Launch date not confirmedDownloadable