Update README.md

README.md

@@ -21,14 +21,14 @@ models/unet directory.
 [original_model_name]_mxN_N.gguf
 
 where N_N is the average number of bits per parameter.
+
+## Good choices to start with
 ```
--  9_6 might just fit on a 16GB card
--  9_2 (new) might be better for 16GB cards
--  8_4 is a good balance for 16GB cards,
--  7_4 is roughly the size of an 8 bit model,
--  5_9 should work for 12 GB cards
--  5_1 is mostly quantised to Q4_1 
+-  9_2 might be best for 16GB cards
+-  7_6 or 7_4 might work for 12 GB cards
+-  5_9 should definitely work for 12 GB cards
 ```
+
 ## How is this optimised?
 
 The process for optimisation is as follows: