Alexander Hanff on Nostr: Qwen2.5-coder:32b has dropped and it is jaw dropingly good - beats OpenAi's GPT-4o ...

Qwen2.5-coder:32b has dropped and it is jaw dropingly good - beats OpenAi's GPT-4o and is on par with Claude 3.5 Sonnet (the highest benchmarked model for coding to date).

I got it to generate an entire blogging platform with MFA, threaded comments, categories, front and backend, mongodb storage and bootstrap/jquery front end - in ~5 minutes on Mac Studio M1 Ultra with 128GB of RAM via Ollama with a context window of 96k... astonishingly good.