LocalLLaMA @sh.itjust.works noneabove1182 @sh.itjust.works 2y ago WizardLM introduce the newest WizardCoder 34B based on Code Llama twitter.com /WizardLM_AI/status/1695396881218859374 ✅WizardCoder-34B surpasses GPT-4, ChatGPT-3.5 and Claude-2 on HumanEval with 73.2% pass@1
Cool, but comparison is a stretch, as admitted by the authors. With identical test methodology gpt-4 is still better Still a good news
Cool, but comparison is a stretch, as admitted by the authors. With identical test methodology gpt-4 is still better
Still a good news
Agreed, but still huge progress in OSS models in a very short time!