Granite 3.2 8B Instruct
Granite 3.2 8B Instruct adds controllable thinking to Granite 3.1 8B Instruct. Released on 26 February 2025, it remains downloadable and listed for dedicated watsonx deployment.
Release details
The model is post-trained from an existing instruction checkpoint. Training combines permissively licensed public data and synthetic reasoning examples. A chat-template option enables or disables thinking. Supported tasks include retrieval-based answers, extraction, coding and function calls.Granite 3.2 8B Instruct model card
Practical use
Practical guidance. Keep this release when reproducing earlier results or maintaining a specific integration. For new work, compare it with the later Granite models on the required behavior. Availability of older weights does not mean they are the best default for a new application.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API, Downloadable, Provider product |
| Model identifiers | ibm-granite/granite-3.2-8b-instruct |
| License | Apache 2.0Granite 3.2 8B Instruct model card |
| Parameters | 8BGranite 3.2 8B Instruct model card |
| Successor identified by the card | Granite 3.3 8B InstructGranite 3.2 8B Instruct model card |
| Hosted access | watsonx: deploy on demand.IBM watsonx.ai foundation model library |
Related models
- Granite 4.2 8B
IBM · Granite 4.2
An 8B dense text model derived from Granite 4.1 Base with full, low-effort and direct-answer thinking modes.
Launch date Downloadable - Granite 3.3 8B Instruct
IBM · Granite 3.3
An 8B instruction model with optional structured reasoning and support for 12 languages.
Launch date API · Downloadable · Provider product