My local models are worse than Claude at almost everything, but they still run most of my workflow for $0

Local AI models may not be as advanced as Claude in performing certain tasks, but they still serve the majority of my workflow without any cost. For instance, when I need to analyze spending reports or generate bar graphs and flow charts, my local AI stack on my RTX 4070 Ti Super is more than capable. Even when it comes to parsing documents, Gemma 4 E4B, with its vision capabilities and large context size, is well-suited for the job.

While Claude might occasionally feel restrictive due to usage limits, using a local model allows me to maintain continuity without breaking the reasoning trail. In coding sessions, I often let my Qwen3:30B MoE model generate a working Python app first, then have Claude Code refine it, saving me a significant amount of money. When it comes to non-work-related tasks like recommending games, books, or shows, local models like Gemma 4 can even outperform Claude in providing better suggestions.

Ultimately, while frontier AI models excel in complex tasks, for everyday needs, my local AI stack offers a more cost-effective and practical solution.