jundot/omlx v0.7.3 was released on October 31, 2026. The release information covers improvements to DeepSeek V4.1 handling, the voice conversation and web search features, and memory usage across multiple requests. The source of these changes is the v0.7.3 release of jundot/omlx. The core of this update therefore spans specific model handling, conversation features, and memory usage when dealing with multiple requests.
jundot/omlx is a repository for an LLM inference server for Apple Silicon. The repository description states that it supports continuous batching and an SSD cache and is managed from the macOS menu bar. This release description lists DeepSeek V4.1 handling, voice conversation, web search, and memory usage across multiple requests together as improvements to that server. However, the specific implementation approach for each item does not appear in the given release summary. Accordingly, the verifiable scope extends only as far as which parts were targeted for improvement.
The release summary presents no performance figures, comparison baselines, or validation conditions. So how much DeepSeek V4.1 handling has changed, and how large the improvement in memory usage is, cannot be judged from this material alone. Voice conversation and web search are only mentioned as features, and the release summary does not clarify how to use them or what is provided. v0.7.3 is an update whose improvement items can be checked, but the actual effects and detailed behavior are hard to determine from the release summary alone.