![](https://cdn-avatars.huggingface.co/v1/production/uploads/6353e36a5eac2d2efa75699e/r2jeIt9rK2ggVfUwPqoa6.png)
legraphista/gemma-2-2b-it-IMat-GGUF
Text Generation
•
Updated
•
295
•
1
These quantized models have a smaller memory footprint, but acceptable quality.
Note Claims to outperform GPT-3.5. MMLU Pro scores of the Open LLM Leaderboard puts it slightly below OpenHermes 2.5. 3 GB RAM: Go for IQ3_M quant (tested with my phone)
Note Outperforms Gemma-2 2B in terms of MMLU Pro score (33.58% versus 17.22% according to the Open LLM Leaderboard), but is slightly bigger at 3.8B parameters vs. 2.6B, and not so great at roleplay.
Note My goto "jack-of-all-trades" model, whether it be my studying or roleplaying partner.