Different results from LM Studio and Ollama with same local model (read_file)

[Sab3rRid3r]

When I use Ollama to request information about content from the project using local models, access is not granted correctly. A message is displayed stating that access is to be granted, but this does not happen.

When using LM Studio, a “Continue read” (read_file) always occurs and access works.

I tested this with Ollama on Linux and macOS and LM Studio on macOS and Windows.

We have a Linux server with the appropriate GPU in our company that provides models via Ollama. It would be nice to find out why it works worse with Ollama than with LM Studio.

Thanks in advance.

Model qwen3-coder-30b 4bit (has tool_use capabilities)
Continue 1.0.60
LM Studio 0.4.5+2
Ollama 0.17.0

• name: Qwen3-coder-30b lmstudio
  provider: lmstudio
  model: qwen/qwen3-coder-30b
  defaultCompletionOptions:
  contextLength: 64000
  temperature: 0.2
  topP: 0.95
  topK: 40
• name: Qwen3-coder-30b ollama
  provider: ollama
  model: qwen3-coder:30b
  defaultCompletionOptions:
  contextLength: 256000
  maxTokens: 20000
  temperature: 0.2
  topP: 0.95
  topK: 40