Gemma 2 27b running locally on my laptop sometimes gives me far better results than gpt-4o, while being almost as fast. It's an incredible #LLM for its comparatively small size.