Models and responses
First verify Settings → Server → Test Connection succeeds. If it fails, use the connection checklist before changing models.
A model is missing
Section titled “A model is missing”Confirm it is downloaded on the configured LM Studio host. Clear the model picker’s search and Format or Capability filters, then refresh. A model on another computer will not appear merely because it was used in an imported chat.
If the conversation says Choose Model, select an available chat model. This can happen when a saved model is missing or unloaded.
A model will not load
Section titled “A model will not load”Check the model in LM Studio directly. The host may lack memory, a compatible runtime, or a complete download. Try a smaller chat model and reduce other host workloads.
Unload Previous Chat Model can reduce overlapping model memory. Use Unload All Models only when it is appropriate to affect every client of the server; see Utilities.
A response stops or is interrupted
Section titled “A response stops or is interrupted”Check whether the host slept, the network disconnected, the app lost background time, or a tool failed. Reopen the chat and inspect any saved partial output and error notice. If the interface offers a recovery action, read its explanation before continuing.
An interrupted response may have partial reasoning or tool output as well as answer text. Do not assume a tool action was rolled back just because the response stopped.
The conversation reaches its context limit
Section titled “The conversation reaches its context limit”Inspect Context and model details beside the composer. The loaded context length may be lower than the model’s maximum.
Start a new chat with a concise summary of what matters, reduce large inputs, or adjust the model’s loaded context on the host if its memory permits. The context sheet reports usage; it does not allocate more host memory.
An imported chat asks about the server
Section titled “An imported chat asks about the server”Saved continuation metadata belongs to the server context that created it. Select Same LM Studio Server only if that is true. Use Transcript Only during import or Start New Context when prompted in a chat if you cannot reuse the saved context. The visible transcript and server-side continuation are different things.
Reasoning is not visible
Section titled “Reasoning is not visible”Check the model’s reasoning capability and Stream Reasoning. With streaming disabled, expand the completed reasoning section after the reply. The separate Reasoning generation switch only affects compatible models and new requests.
If the problem persists, record the model name, app and LM Studio versions, and error text for Support.