Diff
1diff --git a/dot_config/llama-server/models.ini b/dot_config/llama-server/models.ini
2index 4f1bdc868fa5cc2808a194e20ccb50bdfd9ad500..4bdacf387ec534215ff40ed1e323414c1b7db898 100644
3--- a/dot_config/llama-server/models.ini
4+++ b/dot_config/llama-server/models.ini
5@@ -1,12 +1,13 @@
6-[embeddinggemma-300m]
7+[ggml-org/embeddinggemma-300M-GGUF:Q8_0]
8 hf = ggml-org/embeddinggemma-300M-GGUF:Q8_0
9 alias = embeddinggemma-300m
10 n-gpu-layers = 99
11 ctx-size = 2048
12 embedding = true
13
14-[ministral3-8b]
15+[unsloth/Ministral-3-8B-Instruct-2512-GGUF:Q4_K_XL]
16 hf = unsloth/Ministral-3-8B-Instruct-2512-GGUF:UD-Q4_K_XL
17+alias = ministral3-8b
18 n-gpu-layers = 99
19 ctx-size = 16384
20 flash-attn = on
21@@ -14,7 +15,7 @@ jinja = true
22 temp = 0.15
23 parallel = 1
24
25-[qwen3.6-27b]
26+[unsloth/Qwen3.6-27B-MTP-GGUF:Q4_K_M]
27 hf = unsloth/Qwen3.6-27B-MTP-GGUF:Q4_K_M
28 alias = qwen3.6-27b
29 n-gpu-layers = 99
30@@ -31,7 +32,7 @@ spec-draft-n-max = 2
31 repeat-penalty = 1.0
32 reasoning = on
33
34-[qwen3.6-35b-a3b]
35+[unsloth/Qwen3.6-35B-A3B-GGUF:Q4_K_S]
36 hf = unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q4_K_S
37 alias = qwen3.6-35b-a3b
38 n-gpu-layers = 99
39@@ -46,25 +47,25 @@ parallel = 1
40 repeat-penalty = 1.0
41 reasoning = on
42
43-[gemma4-26b]
44+[unsloth/gemma-4-26B-A4B-it-qat-GGUF:Q4_K_XL]
45 hf = unsloth/gemma-4-26B-A4B-it-qat-GGUF:UD-Q4_K_XL
46-alias = gemma4-26B
47+alias = gemma4-26b-a4b
48 n-gpu-layers = 99
49 n-gpu-layers-draft = 99
50 ngl=999
51 flash-attn = on
52 ctx-size = 131072
53-temp = 1.0
54-top-p = 0.95
55-top-k = 64
56+temp = 0.85
57+top-p = 0.9
58+top-k = 32
59 spec-type = draft-mtp
60 spec-draft-n-max = 4
61 fit = off
62 reasoning = auto
63
64-[muse-glimmer-30b]
65+[unsloth/Muse-Glimmer-30B-GGUF:Q4_K_XL]
66 hf = unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL
67-alias = muse-glimmer
68+alias = muse-glimmer-30b
69 n-gpu-layers = 99
70 flash-attn = on
71 jinja = true