IBM has released Granite Switch 4.1, a new family of models designed to handle chat, decision-making, and speech processing from a single endpoint. These models integrate specialized adapters, including a general-purpose Decision aLoRA and a Guardian aLoRA that outperforms existing decision models on safety tasks. Additionally, Granite Switch 4.1 incorporates Granite-Speech-5.0-Turbo-CTC, an ASR model that leads the OpenASR leaderboard in speed. The architecture allows for efficient reuse of the KV cache, enabling rapid switching between different task types and offering enhanced performance with reduced VRAM requirements. AI
IMPACT Consolidates multiple AI capabilities into a single endpoint, potentially reducing hardware costs and increasing inference speed for diverse applications.
RANK_REASON IBM's release of a new model family with integrated capabilities. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Clef-Flash
- Decision aLoRA
- GPT-Luna-Decisions
- Granite Libraries
- Granite Switch 4.1
- Granite-Switch-4.1-30B
- Granite-Switch-4.1-8B
- Guardian aLoRA
- Jevíčko
- Microsoft-Decision-1
- OpenASR leaderboard
- Qwen
- vLLM
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
