Falcon3 3B Instruct
Falcon3 3B Instruct provides multilingual text generation in a 32K-context transformer. Its base was obtained by pruning Falcon3-7B and retraining the reduced model.
How this Falcon3 size was made
The model has 22 decoder blocks. TII reports 100 billion tokens of training for the pruning-and-recovery stage, followed by instruction post-training. Its four stated languages are English, French, Spanish and Portuguese.Falcon3-3B-Instruct model card
Compare reduced models on difficult fields
Practical guidance. A compact extracted record can look correct while omitting an uncommon name or qualifier. Use examples with such details and score missing fields separately. This helps establish whether the reduced model is adequate for the intended workflow.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | December 2024 |
| Access | Downloadable |
| Model identifiers | tiiuae/Falcon3-3B-InstructFalcon3-3B-Base |
| Context window | 32,768 tokens Instruction checkpoint limit, corroborated by its max_position_embeddings configuration. Falcon3-3B-Instruct model card |
| License | TII Falcon-LLM License 2.0Falcon3-3B-Instruct model card |
| Weight access | Weights remain published.Falcon3-3B-Instruct model card |
| Variants | Base and Instruct; this page describes InstructFalcon3-3B-Instruct model card |
Related models
A 7B transformer language model for English, French, Spanish and Portuguese with a 32K context.
Launch date December 2024Downloadable