I'm not paying $20 for ChatGPT or Claude because a free local LLM does
a day ago
- The author used both ChatGPT and Claude subscriptions costing $20 each per month, but found a local AI model Qwen 3.8-27B that handles most tasks for free.
- Qwen 3.8-27B runs on local hardware (e.g., 4070 Ti Super), achieves 33.7 tokens per second, and can perform tasks like building a Snake game and text summarization.
- ChatGPT tasks such as grammar checking and sentiment analysis are easily handled by Qwen, replacing that subscription.
- Claude is better for coding but has restrictive five-hour usage limits; a local model like Qwen has no such limits and allows uninterrupted iterative work.
- Running a local model means no data leaves the PC, no usage caps, and no additional costs beyond hardware power consumption.
- The author concludes that Qwen 3.8-27B fits their workflow perfectly, and there are other free models on Hugging Face for different hardware.