There is also Phind and WizardLM, but I found their performance to be quite poor tbh. It's hard to match gpt-3.5 quality.