In Claude Desktop, you can now keep the familiar Anthropic interface, but replace the model that actually responds to requests. The open-source Ollama platform has made it possible to connect local and cloud models from its catalog, including Qwen, DeepSeek, Kimi, and GLM.
In Claude Desktop, you can now keep the familiar Anthropic interface, but replace the model that actually responds to requests. The open-source Ollama platform has made it possible to connect local and cloud models from its catalog, including Qwen, DeepSeek, Kimi, and GLM.
This is reported by The New Stack.
Ollama is a platform that allows you to run large language models locally on your computer or via the cloud. It works like this: the user enables the integration with Claude in Ollama, after which the service configures Claude Desktop so that queries can be passed to the models via Ollama.
Why run these models on Claude Desktop? The developers explain that using different models in one application can be more convenient than switching between separate services each time.
This is Ollama’s second attempt to run third-party models inside Claude Desktop. In the spring, the company had already added a command that configured the Anthropic application to work through Ollama. But then Claude Desktop started rejecting model identifiers that were not from Anthropic. So Ollama removed this feature.
And now, in Ollama version 0.33.0, a separate local proxy for Claude Desktop has appeared. So, Claude Desktop connects to Ollama as an intermediary, and Ollama already passes requests to the selected model.
Previously, dev.ua wrote about Google’s open Gemma 4 model, which can be run locally, in particular via Ollama.
