LM Studio’s Bionic 1.1.10 update brings an in-app terminal to public builds and adds support for more local-model workflows, including speculative decoding for MLX models. The release also adds an oMLX engine for Apple Silicon Macs and makes several everyday tasks less fiddly.
LM Studio Watch analysis
What happened
The update, dated 29 October, gives Bionic users an in-app terminal with configurable shell profiles and restored terminal sessions that start automatically when selected. LM Studio’s changelog also lists Excalidraw previews in chat, faster project loading and on-device Read Aloud for Linux and Windows, as well as Mac.
For local model users, the release adds SafeTensors support for speculative decoding, speculative decoding for MLX models, and embedding and decision models in the catalogue. It also offers additional compatible LLM backend downloads in Runtime settings, an experimental GPU cache setting for llama.cpp MoE models, and improved runtime compatibility and download selection for M5 Macs. These are changes listed in LM Studio’s Bionic 1.1.10 changelog.
Our top picks
- In-app terminal
Public Bionic builds now include a configurable terminal, with screen-reader support and saved sessions that restart when selected. - More speculative decoding options
The release adds support for SafeTensors models and MLX models, broadening the local setups that can use the technique. - More model and backend choices
Embedding and decision models join the catalogue, alongside additional compatible backend downloads in Runtime settings. - Read Aloud beyond Mac
On-device Read Aloud now also works on Linux and Windows, with seven additional offline voices. - Less friction in long sessions
LM Studio says it has improved context compaction when conversations exceed model limits and reduced memory use during large session exports and moves. - An experimental llama.cpp option
A new GPU cache setting for llama.cpp mixture-of-experts models gives users another setting to explore, though it is explicitly experimental.
Why it matters
This is a broad update to a desktop tool for running models locally, not just a fresh coat of interface paint. The terminal, backend downloads and wider model support give users more ways to work with local setups from inside the app; Read Aloud and session-management changes make that setup more useful day to day.
The features are not all equally mature. In particular, the llama.cpp GPU cache option is experimental, so it belongs in the “try carefully” column rather than the “settled feature” one.
Our read
The best part is the practical spread: LM Studio is adding capabilities for developers, model tinkerers and people who simply want their local assistant to read something aloud. An in-app terminal is a substantial addition, and the extra model and backend options make the release more than a seasonal theme with a changelog attached.
If you use Bionic, the changelog is worth a scan for the features that match your setup. Treat the experimental cache setting as an experiment, not a promise.
What to watch
- Whether the new terminal and restored sessions behave reliably across supported platforms.
- Which models and backends work well with the added speculative-decoding options.
- Whether the experimental llama.cpp GPU cache setting graduates into a standard feature.
Discussion spark: Should a local AI app like LM Studio keep adding developer tools such as an in-app terminal, or focus on making model setup and everyday use simpler?
Sources and evidence
- Bionic Changelog • 1.1.10 | LM Studio (29 October 2026, 00:00 UTC)
not affiliated with or endorsed by LM Studio or Element Labs