@bendyline/gilde 0.1.33 → 0.1.34

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,194 @@
1
+ {
2
+ "schemaVersion": 1,
3
+ "kind": "chat-model",
4
+ "id": "qwen3.8-27b-q3",
5
+ "name": "Qwen 3.8 (27B, Q3)",
6
+ "description": "Alibaba's Qwen 3.8 at 27B parameters: a dense, multimodal agentic model with native text, image, and video understanding, 262K context, flexible reasoning effort, MTP, and native tool calling. Unsloth's dynamic UD-Q3_K_XL is the smallest build in this line — smaller on disk than a plain Q3_K_M and better preserved at the same bit width — bringing the 27B within reach of machines with 20GB+ usable memory.",
7
+ "tags": [
8
+ "alibaba",
9
+ "agentic",
10
+ "coding",
11
+ "multimodal",
12
+ "vision",
13
+ "video",
14
+ "tools",
15
+ "reasoning",
16
+ "long-context"
17
+ ],
18
+ "category": "general",
19
+ "maintainer": {
20
+ "name": "Alibaba",
21
+ "url": "https://huggingface.co/Qwen/Qwen3.8-27B"
22
+ },
23
+ "version": "1.0.0",
24
+ "updatedAt": "2026-08-17T00:00:00Z",
25
+ "license": "Apache-2.0",
26
+ "licenseClass": "open",
27
+ "licenseShortName": "Apache 2.0",
28
+ "recoScore": 20,
29
+ "parameterSize": "27B",
30
+ "approxSizeBytes": 13441059904,
31
+ "supportsTools": true,
32
+ "contextWindow": 262144,
33
+ "upstream": "https://huggingface.co/Qwen/Qwen3.8-27B",
34
+ "style": {
35
+ "family": "qwen",
36
+ "reasoningFormat": "think",
37
+ "toolCallFormat": "function-call"
38
+ },
39
+ "behaviors": [
40
+ "reasoning.strip-think-tags",
41
+ "provider.merge-system-messages",
42
+ "mcp.compact-tool-schemas",
43
+ "provider.compact-write-transcript",
44
+ "fabrication.detect-past-tense-no-tools",
45
+ "turn.ollama-num-predict-bumped",
46
+ "turn.preamble-folding",
47
+ "turn.ramble-detection",
48
+ "tools.mlx-grammar",
49
+ "prompt.retrieval-first",
50
+ "prompt.workspace-gestalt",
51
+ "prompt.derive-by-execution"
52
+ ],
53
+ "tuning": {
54
+ "sampling": {
55
+ "temperature": 1,
56
+ "topP": 0.95,
57
+ "topK": 20,
58
+ "minP": 0,
59
+ "presencePenalty": 0,
60
+ "repetitionPenalty": 1,
61
+ "maxTokens": 12288
62
+ },
63
+ "samplingWhenThinking": {
64
+ "temperature": 1,
65
+ "topP": 0.95
66
+ },
67
+ "reasoning": {
68
+ "enableThinking": true,
69
+ "thinkingBudget": 4096,
70
+ "templateKwargs": {
71
+ "reasoning_effort": "medium"
72
+ }
73
+ },
74
+ "promptTags": {
75
+ "enableThinkingTag": "/think",
76
+ "disableThinkingTag": "/no_think"
77
+ },
78
+ "profiles": {
79
+ "thinking-general": {
80
+ "sampling": {
81
+ "temperature": 1,
82
+ "topP": 0.95,
83
+ "topK": 20,
84
+ "minP": 0,
85
+ "presencePenalty": 0,
86
+ "repetitionPenalty": 1,
87
+ "maxTokens": 12288
88
+ },
89
+ "reasoning": {
90
+ "enableThinking": true,
91
+ "thinkingBudget": 4096,
92
+ "templateKwargs": {
93
+ "reasoning_effort": "medium"
94
+ }
95
+ }
96
+ },
97
+ "thinking-deep": {
98
+ "reasoning": {
99
+ "enableThinking": true,
100
+ "thinkingBudget": 4096,
101
+ "templateKwargs": {
102
+ "reasoning_effort": "xhigh"
103
+ }
104
+ }
105
+ },
106
+ "thinking-coding": {
107
+ "sampling": {
108
+ "temperature": 1,
109
+ "topP": 0.95,
110
+ "topK": 20,
111
+ "minP": 0,
112
+ "presencePenalty": 0,
113
+ "repetitionPenalty": 1,
114
+ "maxTokens": 16384
115
+ },
116
+ "reasoning": {
117
+ "enableThinking": true,
118
+ "thinkingBudget": 4096,
119
+ "templateKwargs": {
120
+ "reasoning_effort": "medium"
121
+ }
122
+ }
123
+ },
124
+ "thinking-precise": {
125
+ "sampling": {
126
+ "temperature": 0.6,
127
+ "topP": 0.95,
128
+ "topK": 20,
129
+ "minP": 0,
130
+ "presencePenalty": 0,
131
+ "repetitionPenalty": 1,
132
+ "seed": 0,
133
+ "maxTokens": 6144
134
+ },
135
+ "reasoning": {
136
+ "enableThinking": true,
137
+ "thinkingBudget": 4096,
138
+ "templateKwargs": {
139
+ "reasoning_effort": "medium"
140
+ }
141
+ }
142
+ },
143
+ "instruct": {
144
+ "sampling": {
145
+ "temperature": 0.7,
146
+ "topP": 0.8,
147
+ "topK": 20,
148
+ "minP": 0,
149
+ "presencePenalty": 1.5,
150
+ "repetitionPenalty": 1,
151
+ "maxTokens": 4096
152
+ },
153
+ "reasoning": {
154
+ "enableThinking": false
155
+ }
156
+ },
157
+ "creative": {
158
+ "sampling": {
159
+ "temperature": 1,
160
+ "topP": 0.95,
161
+ "topK": 40,
162
+ "minP": 0,
163
+ "presencePenalty": 1.5,
164
+ "repetitionPenalty": 1,
165
+ "maxTokens": 8192
166
+ },
167
+ "reasoning": {
168
+ "enableThinking": false
169
+ }
170
+ }
171
+ },
172
+ "engine": {
173
+ "llamaCpp": {
174
+ "spec": {
175
+ "mtp": true,
176
+ "nMax": 4
177
+ }
178
+ }
179
+ }
180
+ },
181
+ "llamaCpp": {
182
+ "huggingfaceRepo": "unsloth/Qwen3.8-27B-GGUF",
183
+ "revision": "f1bfb127c64f7072bdd2cad55f258b9c8b2910fe",
184
+ "filename": "Qwen3.8-27B-UD-Q3_K_XL.gguf",
185
+ "sha256": "00cf92e666c6af6566996c38c89a44ccdb6449ea25ef0f112a452c853b2a71e2",
186
+ "approxSizeBytes": 13441059904,
187
+ "quantization": "UD-Q3_K_XL",
188
+ "mmproj": {
189
+ "filename": "mmproj-F16.gguf",
190
+ "sha256": "cbb841a9ee0636b2ec172f5bb8df2ea8dfeb01e90fe7c6126581d662a0b4e43e",
191
+ "sizeBytes": 927607488
192
+ }
193
+ }
194
+ }
@@ -0,0 +1,19 @@
1
+ {
2
+ "schemaVersion": 1,
3
+ "version": "1.0.0",
4
+ "releasedAt": "2026-08-17T00:00:00Z",
5
+ "approxSizeBytes": 13441059904,
6
+ "llamaCpp": {
7
+ "huggingfaceRepo": "unsloth/Qwen3.8-27B-GGUF",
8
+ "revision": "f1bfb127c64f7072bdd2cad55f258b9c8b2910fe",
9
+ "filename": "Qwen3.8-27B-UD-Q3_K_XL.gguf",
10
+ "sha256": "00cf92e666c6af6566996c38c89a44ccdb6449ea25ef0f112a452c853b2a71e2",
11
+ "approxSizeBytes": 13441059904,
12
+ "quantization": "UD-Q3_K_XL",
13
+ "mmproj": {
14
+ "filename": "mmproj-F16.gguf",
15
+ "sha256": "cbb841a9ee0636b2ec172f5bb8df2ea8dfeb01e90fe7c6126581d662a0b4e43e",
16
+ "sizeBytes": 927607488
17
+ }
18
+ }
19
+ }