IBM has released Granite Switch 4.1.1, a patch update to its Granite Switch 4.1 model family, published on Hugging Face. Released just over a week after version 4.1, the update is fully backwards compatible, requires no migration, and contains incremental improvements. Two components changed. The Granite Speech automatic speech recognition model now transcribes ten languages and adds native punctuation and capitalization, so transcripts need no separate formatting step. A revised Decision adapter in granitelib-core, which IBM says performs better on the Jev Decision Index test set, also ships in this release. The authors highlight the speed of the release as the notable part. Apart from training the new speech model and the Decision adapter, nothing was recomputed, and the base model was left untouched. Granite Switch uses a modular architecture in which specialised, cache-efficient adapters are composed onto a single endpoint, so a new adapter can be deployed without retraining the whole model. IBM says the same pattern could support more than patch releases of its own models. It is working on tutorials and an Adapter SDK so developers and third-party providers can build, version and ship their own adapters. Beyond promising "more soon," the post gives no timeline and no benchmark figures for the improvements. For Australian businesses and professionals, the direct impact is limited. Organisations already running Granite Switch 4.1 can likely adopt 4.1.1 without changes to their setup, and teams that transcribe multilingual audio may want to test the new speech model, since native punctuation could reduce post-processing work. Organisations that do not use these models can ignore the update. Any use of speech-to-text on recordings of customers or staff still falls under the Privacy Act 1988 and the Australian Privacy Principles, including where audio is processed by a model hosted offshore, so check data handling before deployment.