A Reddit user is seeking hardware recommendations for running multiple small to medium-sized models locally for data parsing, extraction, and reasoning tasks. The user intends to use the setup for model building, testing, LoRA creation, and distillation, while reserving large cloud models like Opus for complex tasks.
- Goal: Run multiple small to medium models locally for data parsing, extraction, and slight reasoning capabilities.
- Additional features: Image generation and computer use are desired but secondary.
- Development focus: The user wants to utilize DGX Sparks for their marketed value in building and testing models locally, rather than just inference.
- Advanced tasks: Interest in building LoRAs and distilling medium-sized models into specific domains.
- Cloud fallback: Large models like Opus will still be used for huge design and difficult bug hunting tasks.