Salamandra 7B FC 2607
Salamandra 7B FC 2607 adapts BSC's 2606 instruction model for function calling in a smaller model class.
Restricted access
Model and training
The card identifies Salamandra 7B Instruct 2606 as its parent. It describes supervised and preference-training stages, but its alignment warning and later text are inconsistent. Its architecture table also contains 40B-sized values. This record therefore omits a numeric context claim. The main weights need manual approval; the official GGUF copy also requires access conditions.Salamandra 7B FC 2607: official model card
Practical use
Practical guidance. Use this release to compare tool reliability at a smaller serving size. Treat the card's alignment claims as unresolved rather than assuming the preference stage guarantees safe behaviour. Inspect the released configuration after access is granted. Compare its function arguments and multi-step outcomes with the 40B FC model.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Restricted access |
| Access | Downloadable |
| Model identifiers | salamandra-7b-fc-2607salamandra-7b-fc-2607-GGUF |
| License | Apache License 2.0.Salamandra 7B FC 2607: official model card |
| Model size | 7B model class; published architecture table is inconsistent.Salamandra 7B FC 2607: official model card |
Related models
Reasoning and function-calling model with 40 billion parameters and access-controlled weights.
Launch date not confirmedRestricted accessDownloadableSalamandra 7B Instruct 2606 updates BSC's conversational model with longer context and preliminary preference alignment.
Launch date not confirmedDownloadable