[Havoc]

3090 or 4090 will be able to run quantized 22B models.

Though realistically for code completion smaller models will be better due to speed