Granite 3.3 2B Instruct
Granite 3.3 2B Instruct is an April 2025 model with optional structured reasoning. It is adapted from Granite 3.3 2B Base and supports 12 languages.
Release details
The chat template accepts a thinking option. Separate thought and response tags identify intermediate reasoning and the final answer. The model supports extraction, retrieval-based answers, coding and tool calls. Its training combines permissively licensed examples with curated synthetic reasoning tasks.Granite 3.3 2B Instruct model card
Practical use
Practical guidance. The 2B model can provide a small-model baseline for existing Granite 3.3 deployments. Compare thinking enabled and disabled on the same short tasks. Retain the response tags during parsing so reasoning text is not accidentally shown as the final answer.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API, Downloadable, Provider product |
| Model identifiers | ibm-granite/granite-3.3-2b-instruct |
| Context window | 131,072 tokens Published checkpoint context; hosted service limits can differ. Granite 3.3 2B Instruct model card |
| License | Apache 2.0Granite 3.3 2B Instruct model card |
| Parameters | 2BGranite 3.3 2B Instruct model card |
| Hosted access | watsonx: deploy on demand.IBM watsonx.ai foundation model library |
Related models
- Granite 4.2 3B
IBM ยท Granite 4.2
A 3B dense text model derived from Granite 4.1 Base with full, low-effort and direct-answer thinking modes.
Launch date Downloadable