Update README.md

README.md

@@ -64,8 +64,16 @@ Raise/change rep pen SLOWLY ie: 1.02, 1.021, 1.022 etc etc.
 
 Recommend temps under 1 for a wee bit crazy, 1-2 for crazy, and over 2 for way over the top, full on madness.
 
+Suggest start settings:
+
+Temp: .5 to .8 , Rep pen 1.02 to 1.05, Context min: 4k/4096 characters (as this model can output 2k+ at a time)
+
+To limit output length, put limits (ie 800-1000 words) in the prompt AND/OR set a hard limit on output generation.
+
+Definitely set a "hard limit" for role play and/or chat.
+
 Different quants will give you slightly different prose, with higher quants giving strongest level of detail, "there"
-and nuance.
+and nuance. Q4+ recommended.
 
 Q2k: If you use this quant you may need to lower temp (less then 1) and raise rep pen (1.08+) to address quality loss.