Stop using Claude
Cost per task is literally 100x higher than DeepSeek and deepseek even beats opus 5 on some benchmarks
Mfs did all their training on open source code, yet contribute nothing back t...
I’ve had a much better time using Codex than Claude. I don’t know what all the hype around Claude is about. Codex is much faster and more accurate.
I know Codex isn’t open-source either, but I don’t like sharing my data with a Chinese company. If someone has the infrastructure to run it locally, though, maybe it’s a best choice. At the very least, it’s open-source.
My machine can only comfortably support 8B models, which isn’t sufficient for coding purposes or answering any real question. But hopefully more powerful infrastructure becomes more affordable.
I know Codex isn’t open-source either, but I don’t like sharing my data with a Chinese company. If someone has the infrastructure to run it locally, though, maybe it’s a best choice. At the very least, it’s open-source.
My machine can only comfortably support 8B models, which isn’t sufficient for coding purposes or answering any real question. But hopefully more powerful infrastructure becomes more affordable.
31❤️1🚀1
