Nemotron-Mini 4B Instruct

Good fit for

People building local character conversations or small-model retrieval experiments in English.

Why consider it

The runtime documentation explicitly identifies roleplay, RAG question answering and function calling as intended uses.

Main limitations

The documented context window is 4,096 tokens, so long fictional histories need context management. RAG and tools require connected software. A small download does not establish performance on every device.

Cost and access

No hosted subscription for the local package within its permitted uses; hardware and electricity remain your responsibility. NVIDIA AI Foundation Models Community License Agreement applies. The catalog describes commercial readiness, but review the complete terms before production.

Getting started

Install a compatible Ollama version, review the license, then run: ollama run nemotron-mini:4b. API means the runtime-provided local API, not a hosted endpoint.