There’s a new sheriff in LLM land! Please meet Arcee Nova, a new model based on Qwen2–72B. Nova outperforms the best models listed on the Hugging Face LLM leaderboard and closes the gap on GPT-4 performance.
In this video, I run a 3-bit quantized version on my MacBook and ollama, with a couple of creative writing and multi-step reasoning prompts. Excellent results and very decent speed for a model this large!
The future is open 🚀🚀🚀