llamastash/llamastash v0.10.1 is a patch release centered on the recommendation engine. This version reflects 62 currently public-weights models in the catalog and changes the baseline for benchmark scores to the Artificial Analysis Intelligence Index. The changes can be confirmed in the GitHub release published on 2026-10-30.
llamastash is a local LLM orchestrator, described as running multiple execution engines and models behind the same proxy from the terminal. The repository description says that engines or models can run through the same proxy even when they differ, but the actually supported combinations are not in the given information. The two changes in this update correspond respectively to the model list the recommendation engine consults and the criteria used to read scores. The fact that public-weights models were added to the catalog means that list has been refreshed for currently public models, and the figure of 62 is the value stated in the release summary.
However, the provided release information includes neither the individual names nor the scores of the 62 models, nor the score correspondence between the previous criteria and the new ones. Therefore it is not possible to judge from the criteria change alone how the recommendation rankings shifted, or whether recommendation quality improved. What can be confirmed in this version is the update of the catalog and the scoring criteria, and changes in the actual recommendation results cannot be confirmed from the given information alone.