Details
Granite models span multiple modalities: language, code, time series forecasting, speech transcription, vision (document and image understanding), and embedding. IBM also produces Granite Guardian models designed to detect harmful content, bias, hallucinations, and jailbreak attempts in AI systems. The models are available on IBM's watsonx.ai platform and on Hugging Face. IBM provides IP indemnification for enterprise customers using IBM-developed Granite models, and publishes training data documentation as part of its transparency commitment.
Have evidence about IBM's AI practices? Submit a report.
Report a Sighting →