Bloom/gpt neo/codegen is a good one and all come in various sizes, CPUs are slow though vs GPUs even old GPUs.

Also try https://text-generator.io which is a much more cost effective way to prototype new tools/what's possible right now, way cheaper than OpenAi, does text and code and a 100 requests a month free teir, then you can drop down to doing it yourself once your at a large scale.

Text generator runs under a second normally on fast GPUs