ALIA 40B FC 2606
ALIA 40B FC 2606 adds explicit reasoning to BSC's function-calling model line.
Model and training
The release replaces the non-thinking approach of FC 2605 with reasoning-oriented post-training. It retains the context-extended 40B ALIA foundation and supports tool-directed interaction. The published GGUF conversion is the same trained release in another format. The card does not establish a BSC-operated public inference API.ALIA 40B FC 2606: official model card
Practical use
Practical guidance. Assess the whole tool sequence, including unnecessary calls and recovery after a tool returns an error. Reasoning text can consume the available output budget without improving the final action. Compare task success, latency and total generated tokens against FC 2605 using the same tool definitions.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | ALIA-40b-fc-2606ALIA-40b-fc-2606-GGUF |
| Context window | 163,840 tokens Maximum sequence length documented for these weights; the prompt and generated text share the window. ALIA 40B FC 2606: official model card |
| License | Apache License 2.0.ALIA 40B FC 2606: official model card |
| Model size | 40B class.ALIA 40B FC 2606: official model card |
Related models
Non-thinking 40-billion-parameter model trained for function calling.
Launch date not confirmedDownloadableReasoning and function-calling model with 40 billion parameters and access-controlled weights.
Launch date not confirmedRestricted accessDownloadable