IBM open-sources Granite Switch 4.1: one endpoint for chat, decisions and speech
jacek2023 · reddit · 2026-10-16
IBM released Granite Switch 4.1 (Apache 2.0) in 30B and 8B sizes, a modular LLM that composes a base model (128K context), on-demand aLoRA adapters (decision, guardian-core) and a 470M speech-to-text encoder behind a single endpoint, controlled via chat-template control tokens. Backed by Transformers and vLLM. In the demo, a 3B variant plays Doom at 60-100 decisions/second — vs 10/s for the previously reported Jev — while narrating and answering questions. Preview bundles without audio and the granite-switch composition tool are also open-sourced.
More from Models
- Reflection open-sources 501B MoE Beam; vLLM day-0 support hits 708 tokens/s — Xianbao_QIAN · 2026-10-16
- Beam's Ten-Benchmark Breakdown: Doubles Peers on CritPt, Near-Bottom on Terminal-Bench — ArtificialAnlys · 2026-10-16
- Beam Scores 28 on Coding Agent Index, 40 Behind Leaders but Among Most Token-Frugal — ArtificialAnlys · 2026-10-16
- Beam Token Breakdown: Only 26% of Its 26k Output Tokens per Task Are Answers — ArtificialAnlys · 2026-10-16
- Reflection's First Model Beam Scores 30, Ranks #2 Among US Open Weights Models — ArtificialAnlys · 2026-10-16
- Why are sub-15B open models so rare now? Labs seem to have settled on 24-27B minimum — Hot_Example_4456 · 2026-10-16