f84b279c55acd8cd7d34b244e77ba31a6430a2b5

 1diff --git a/dot_config/llama-server/models.ini b/dot_config/llama-server/models.ini
 2index 89430b6ceae31f4cf13996dae8531a271a538f35..9d2bbe62d27de68cd08067e631911ebb384c7caf 100644
 3--- a/dot_config/llama-server/models.ini
 4+++ b/dot_config/llama-server/models.ini
 5@@ -11,17 +11,21 @@ alias = f2llm-v2-330m
 6 n-gpu-layers = 99
 7 embedding = true
 8 
 9-[empero-ai/Qwen3.8-9B-GGUF:Q6_K]
10-hf = empero-ai/Qwen3.8-9B-GGUF:Q6_K
11-alias = qwen3.8-9b,reviewer
12+[ibm-granite/granite-4.2-8b-GGUF:Q6_K]
13+hf = ibm-granite/granite-4.2-8b-GGUF:Q6_K
14+alias = granite4.2-8b,reviewer
15 n-gpu-layers = 99
16-ctx-size = 32768
17-n-predict = 16384
18+ctx-size = 16384
19+n-predict = 4096
20 flash-attn = on
21 jinja = true
22-temp = 0.15
23-top-k = 20
24+temp = 0.2
25+top-p = 0.95
26+seed = 42
27 parallel = 1
28+reasoning = on
29+reasoning-effort = low
30+reasoning-budget = 2048
31 
32 [unsloth/Qwen3.8-27B-GGUF:Q4_K_XL]
33 hf = unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_XL