Falcon3 10B Instruct
Falcon3 10B Instruct expands the depth of Falcon3-7B-Base. It is a 32K-context text model for English, French, Spanish and Portuguese.
How this Falcon3 size was made
TII increased the model to 40 decoder blocks and trained the expanded model on another two trillion tokens. It then applied instruction post-training. The 10B release is a trained depth expansion, not a simple change to an inference setting.Falcon3-10B-Instruct model card
Evaluate the added depth on the same workload
Practical guidance. Compare 7B and 10B with the same precision, prompt and output budget. Include tasks that require several instructions in one request. The larger checkpoint is useful only if its quality gain meets the application's latency and memory needs.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | December 2024 |
| Access | Downloadable |
| Model identifiers | tiiuae/Falcon3-10B-InstructFalcon3-10B-Base |
| Context window | 32,768 tokens Instruction checkpoint limit, corroborated by its max_position_embeddings configuration. Falcon3-10B-Instruct model card |
| License | TII Falcon-LLM License 2.0Falcon3-10B-Instruct model card |
| Weight access | Weights remain published.Falcon3-10B-Instruct model card |
| Variants | Base and Instruct; this page describes InstructFalcon3-10B-Instruct model card |
Related models
A 7B transformer language model for English, French, Spanish and Portuguese with a 32K context.
Launch date December 2024Downloadable