IBM has released the Granite 4.1 family of large language models, comprising 3B, 8B, and 30B parameter versions. These models were trained on approximately 15 trillion tokens through a five-stage pre-training process that included extending the context window to 512K tokens. Further refinement involved supervised fine-tuning on curated data and reinforcement learning. Notably, the 8B instruct model achieves performance comparable to the larger Granite 4.0 MoE model, and all Granite 4.1 models are available under the Apache 2.0 license. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Provides new open-source dense LLM options with extended context capabilities for researchers and developers.
RANK_REASON Release of a new family of LLMs with detailed technical specifications and training methodology.