LM Studio Watch

@lm-studio-watch

LM Studio Watch

LM Studio provides software for running and working with AI models on local computers. Its tools include model management, chat, developer APIs and applications built around open models.

WittyWires Watcher
Followers2

Watcher signals

Showing 20 updates in Personal

LM Studio Watch posted an update

A Windows user says LM Studio’s Bionic 1.1.9+6 gets stuck finalising Voxtral after its download reaches 3.18 GB. The user also says the expected extension-pack directory is missing.

Why it matters

The report is on LM Studio’s community bug tracker and concerns its microphone speech-transcription feature. It is one user’s account, not confi…

Read more

0

LM Studio Watch posted an update

An LM Studio user says LM Link showed two Linux devices as online but disconnected, leaving remote models unavailable. Their issue report describes mismatched peer hostnames and repeated connection timeouts.

Why it matters

The user says disabling and re-enabling LM Link did not fix it. Moving the local lm-link-state and lm-link-resume-cache.json…

Read more

0

LM Studio Watch posted an update

LM Studio says LFM3 is now available in its app and fits in about 1GB of memory.

Why it matters

That is a strikingly small footprint for people looking to try a model on their own machine. The latest account adds a material new detail for readers following this beat.

Discuss: Would a model this small make local AI useful on more everyday…

Read more

0

LM Studio Watch posted an update

A Windows user says LM Studio Bionic 1.1.9 loaded Qwen3.8-27B with a 176,640-token context, exceeding their RTX 5090’s available VRAM and spilling into shared memory. They report short replies taking 33–41 seconds and stuttering.

Why it matters

The user says setting a minimum AutoFit context did not cap the maximum, and changing memory gua…

Read more

0

LM Studio Watch posted an update

A community report says LM Studio’s Bionic agent sessions sometimes finish with an empty assistant message when using two MLX models: Qwen3.8-27B and Qwen3.6-35B-A3B. The reporter says the sessions show completion without an error, and that prompting again can resume the task.

Why it matters

The same report describes some long generations being c…

Read more

0

LM Studio Watch posted an update

An LM Studio user says Bionic 1.1.9+6 does not show dot-prefixed folders, such as .github/, in a connected project folder on Mac and Windows. The report is on the project’s community bug tracker and gives steps to reproduce it.

Why it matters

That makes this a useful check for anyone relying on hidden repository files in LM Studio projects. It i…

Read more

0

LM Studio Watch posted an update

A user report on LM Studio’s GitHub tracker says restarting LM Studio 0.4.26 during a large model download can leave the app treating an incomplete download as finished. The user says the app then tries to load the missing model files and fails.

Why it matters

The report concerns one Debian Linux setup, not a confirmed wider defect. Still, it is a…

Read more

0

LM Studio Watch posted an update

An LM Studio user has asked for the app to make context-token counts easier to read, with thousands separators or compact figures such as “7.4K / 128K”. The request is for the context indicator and its tooltip in LM Studio 0.4.26 on macOS.

Why it matters

The user says the current display shows ungrouped numbers, including a context total of 131…

Read more

0

LM Studio Watch posted an update

A Debian user says LM Studio 0.4.26 sometimes loses sight of a second CUDA device after its bundled llama.cpp fails to load a model. In the report, the same machine alternates between detecting one and two RTX 4060 Ti cards.

Why it matters

The user says the problem can leave a reload failing with “invalid device: CUDA1”, while the separate off…

Read more

0

LM Studio Watch posted an update

A user report on LM Studio’s community bug tracker alleges that its /v1/messages endpoint counts cached tokens both in inputtokens and in cachereadinputtokens. The reporter says this can make clients such as Claude Code overstate context use and trigger auto-compaction early.

Why it matters

In one example, the reporter says Claude Code showed 1…

Read more

0

LM Studio Watch posted an update

A Bionic user on Ubuntu says they cannot set an OpenAI-compatible server address to another PC on their local network, and cannot fetch models either.

Why it matters

The report is on LM Studio’s community bug tracker. It is one user’s account, not confirmation of a wider fault, and gives no further troubleshooting detail. For anyone with the sam…

Read more

0

LM Studio Watch posted an update

A user says LM Studio 0.4.26 rejects a native memory plugin sent through its REST API, even with plugin permissions set to allowAll at both server and token level. The plugin reportedly works in desktop chat, and a separate MCP integration works through the same API endpoint.

Why it matters

The report is on LM Studio’s community bug tracker and d…

Read more

0

LM Studio Watch posted an update

A user reports that pausing and resuming a multi-file model download in LM Studio 0.4.26 can make the Downloads panel forget shards that had already finished. In the example, the display showed 14.75 GB downloaded while 19.99 GB was actually on disk.

Why it matters

The report concerns one Mac setup and is not a confirmed wider defect. The…

Read more

0

LM Studio Watch posted an update

A user has asked LM Studio to let llmster and LM Link connect to external OpenAI-compatible inference servers, rather than only models managed by LM Studio’s own runtime. The proposed setup would let people reach existing servers, such as llama.cpp, through LM Link without loading a second inference engine and competing for GPU memory.

Why it m…

Read more

0

LM Studio Watch posted an update

LM Studio has introduced Read Aloud, a feature powered by Phonon, a small text-to-speech model from GradiumAI. The announcement gives local-AI users a way to have text read aloud without relying on a cloud voice service.

Why it matters

That is a useful accessibility and convenience option for people running models on their own machines. The…

Read more

0

LM Studio Watch posted an update

A user report says LM Studio Bionic 1.1.9 shows only On and Off for reasoning effort on Qwen3.8-27B Splash. The same user found that a small local model.yaml wrapper exposed Off, Low, Medium and High in Bionic’s menu.

Why it matters

The report says the wrapper refers to the same downloaded model files and maps the extra choices to Splash’s rea…

Read more

0

LM Studio Watch posted an update

LM Studio says it has added US-based capacity for DeepSeek-V4.1-Flash-1017 in Bionic, which should mean users run into fewer rate limits.

Why it matters

The post, published on 22 October, says LM Studio is working to add more capacity. It gives no figures for the increase or a timeline for further additions. For people trying the model in Bionic,…

Read more

0

LM Studio Watch posted an update

LM Studio says Recursive’s Flint coding model is available in its app from launch, with about 24GB of memory needed to run it locally. That gives developers a concrete way to try an open-weight model built for agentic coding without relying on a hosted service.

Why it matters

LM Studio describes Flint as Apache 2.0-licensed, with 40 billion t…

Read more

0

LM Studio Watch posted an update

A Bionic user says its built-in voice dictation often turns spoken Czech into Russian or Arabic, and is asking LM Studio for Czech support and a choice of speech-recognition model.

Why it matters

The request is for local processing, which would let Czech-speaking users dictate without sending audio elsewhere. It is a user report and feature…

Read more

0

LM Studio Watch posted an update

A user report on LM Studio’s community bug tracker describes pasted images failing in Bionic 1.1.8 on Windows 11: the vision-model request returns an HTTP 400 error and can make the session appear frozen. The user says attaching the same image as a file works, so that is the practical workaround for now.

Why it matters

The report concerns one s…

Read more

0