unsloth/alpaca-cleaned
Viewer β’ Updated β’ 51.8k β’ 1.89k β’ 35
An attempt to recreate the feel of GPT-3's 2022 instruct models (text-davinci-002 / text-davinci-003) β direct, unhedged, no assistant persona, no "As an AI language model."
Not the smartest model you'll run. That's sort of the point.
NeoGPT-3 is not affiliated with, endorsed by, or derived from OpenAI's models. It is a QLoRA fine-tune of Mistral-7B-v0.3. "GPT" here describes what it's imitating, not where it came from.