Granite Switch 4.1 8B
Granite Switch 4.1 8B is an open-weight language model from IBM, released on October 14, 2026.
- Developer
- IBM
- Released
- October 14, 2026
- Total parameters
- 8B
- Active parameters per token
- 8B
- Architecture
- Dense
- Context window
- 128K
- Input
- Text, audio
- License
- Apache 2.0
- Commercial use
- Yes
- Official weights, GB
- 17.6 GB
- Hugging Face repository
- huggingface.co
- Checked on
Granite Switch 4.1 8B is a dense model with 8B parameters. It combines the Granite 4.1 8B base model with two adapters (decisions and a safety guardian) and an embedded English speech-to-text encoder, so one endpoint can chat, make decisions and transcribe audio. IBM lists a context window of 131,072 tokens. The model takes text and audio input. The weights are released under the Apache 2.0 license, which allows commercial use.
The official BF16 weights on Hugging Face take about 17.6 GB. Running the model needs at least that much memory across GPUs and system RAM, plus room for the context cache; quantized versions need less.
- Total parameters, billions
- 8B
- Active parameters, billions
- 8B
- Context window, tokens
- 131.07K
- License type
- Apache 2.0
- License conditions
- Apache 2.0: keep the license text and notices; includes a patent grant.
- Weights precision
- BF16
- License text
- apache.org
- Sources
Page URL Hugging Face model card https://huggingface.co/ibm-granite/granite-switch-4.1-8b/raw/main/README.md Hugging Face weights index https://huggingface.co/ibm-granite/granite-switch-4.1-8b/raw/main/model.safetensors.index.json
Questions
What license does Granite Switch 4.1 8B use?
Can I use Granite Switch 4.1 8B commercially?
How much memory does Granite Switch 4.1 8B need?
What context window does Granite Switch 4.1 8B have?
Sources
Every figure comes from the developer’s model card, license text and official announcements, and from the model’s Hugging Face repository metadata. Each model page lists its sources and the date they were checked; the register is rechecked every week.