diff --git a/systems/x86_64-linux/halo/models.ini b/systems/x86_64-linux/halo/models.ini index 00ad6ba..5fca2f8 100644 --- a/systems/x86_64-linux/halo/models.ini +++ b/systems/x86_64-linux/halo/models.ini @@ -41,3 +41,18 @@ parallel = 2 c = 524288 load-on-startup = true chat-template-kwargs = {"preserve_thinking": true} + +[fast] +hf = byteshape/Qwen3.6-35B-A3B-MTP-GGUF:IQ4_XS +spec-type = ngram-simple,draft-mtp +spec-draft-n-max = 3 +threads-batch = 16 +temp = 0.6 +top-p = 0.95 +top-k = 20 +min-p = 0.0 +presence-penalty = 1.5 +parallel = 1 +c = 131072 +chat-template-kwargs = {"preserve_thinking": true} +