Ollama returns HTTP 500

Error message #

llama runner process has terminated: error loading model: unable to locate cpu buffer

What it means #

Ollama could not load the model. Limited memory is one possible cause; the message alone does not identify every cause.

What to try #

1. Try a smaller downloaded model, such as gemma3:1b.
2. Close other memory-intensive applications, then restart Ollama and try a short request.
3. Update Ollama. If the error continues, check its logs and current troubleshooting guidance.

A model’s download size is not its total memory requirement. Larger contexts and other running programs also affect how much memory is available.