@yanauto/broco 0.0.0-stage → 1.0.0
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.
- package/CHANGELOG.md +209 -0
- package/LICENSE +202 -0
- package/README.md +188 -2
- package/README.zh-CN.md +187 -0
- package/assets/broco-board-icon-128.png +0 -0
- package/assets/broco-favicon-192.png +0 -0
- package/assets/broco-favicon-32.png +0 -0
- package/assets/broco-favicon-64.png +0 -0
- package/assets/broco-foreman-256.png +0 -0
- package/assets/broco-hero-1024.png +0 -0
- package/assets/broco-icon-512.png +0 -0
- package/assets/broco-readme-logo-dark.png +0 -0
- package/assets/broco-readme-logo-light.png +0 -0
- package/assets/broco-social-preview-1280x640.png +0 -0
- package/assets/favicon.ico +0 -0
- package/bin/broco +1077 -0
- package/bundle/board/assets/index-Bue0HpFh.css +1 -0
- package/bundle/board/assets/index-DTqMQtD9.js +16 -0
- package/bundle/board/broco-board-icon-128.png +0 -0
- package/bundle/board/broco-favicon-32.png +0 -0
- package/bundle/board/broco-favicon-64.png +0 -0
- package/bundle/board/broco-foreman-256.png +0 -0
- package/bundle/board/favicon.ico +0 -0
- package/bundle/board/fonts/instrument-sans-latin.woff2 +0 -0
- package/bundle/board/fonts/jetbrains-mono-latin.woff2 +0 -0
- package/bundle/board/index.html +18 -0
- package/lib/attachments.mjs +104 -0
- package/lib/claude-md.mjs +132 -0
- package/lib/cli-args.mjs +292 -0
- package/lib/cli-process.mjs +75 -0
- package/lib/client.mjs +198 -0
- package/lib/credential-proxy.mjs +165 -0
- package/lib/credential-source.mjs +23 -0
- package/lib/crew.mjs +947 -0
- package/lib/detect.mjs +279 -0
- package/lib/directories.mjs +32 -0
- package/lib/map.mjs +336 -0
- package/lib/model-catalog.json +864 -0
- package/lib/model-catalog.mjs +138 -0
- package/lib/model-env.mjs +11 -0
- package/lib/native.mjs +228 -0
- package/lib/paths.mjs +63 -0
- package/lib/pi-profile.mjs +29 -0
- package/lib/pi-secrets.mjs +48 -0
- package/lib/pi.mjs +232 -0
- package/lib/platform/index.mjs +13 -0
- package/lib/platform/launchd.mjs +138 -0
- package/lib/platform/linux.mjs +168 -0
- package/lib/platform/macos.mjs +29 -0
- package/lib/platform/processes.mjs +102 -0
- package/lib/problems.mjs +50 -0
- package/lib/projects.mjs +530 -0
- package/lib/quota.mjs +132 -0
- package/lib/roster-output.mjs +136 -0
- package/lib/roster.mjs +913 -0
- package/lib/runner.mjs +403 -0
- package/lib/settings-audit.mjs +104 -0
- package/lib/setup.mjs +416 -0
- package/lib/slots.mjs +177 -0
- package/lib/stats.mjs +92 -0
- package/lib/store.mjs +737 -0
- package/lib/stream-workers.mjs +464 -0
- package/lib/templates.mjs +173 -0
- package/lib/time.mjs +25 -0
- package/lib/web.mjs +260 -0
- package/lib/worker-policy.mjs +17 -0
- package/lib/worktree.mjs +269 -0
- package/package.json +50 -4
- package/scripts/package-board.mjs +53 -0
- package/skills/broco/SKILL.md +59 -0
|
@@ -0,0 +1,864 @@
|
|
|
1
|
+
{
|
|
2
|
+
"as_of": "2026-11-18",
|
|
3
|
+
"source": "Public model documentation, benchmarks, and release notes (Aug–Nov 2026). Not independently verified; grades are relative and change fast.",
|
|
4
|
+
"match_rule": "Match a CLI's model ID to an entry by exact id or alias, after stripping effort/speed suffixes such as -low, -medium, -high, -xhigh, -max, -fast, -thinking.",
|
|
5
|
+
"picks": {
|
|
6
|
+
"hardest_coding_long_tasks": {
|
|
7
|
+
"first": [
|
|
8
|
+
"gpt-6-astra",
|
|
9
|
+
"gemini-4-argon",
|
|
10
|
+
"gemini-4-hydrogen",
|
|
11
|
+
"claude-fable-5"
|
|
12
|
+
],
|
|
13
|
+
"backup": [
|
|
14
|
+
"gpt-6.1-sol",
|
|
15
|
+
"claude-opus-4-6",
|
|
16
|
+
"kimi-k3.1-pro",
|
|
17
|
+
"grok-4.7"
|
|
18
|
+
]
|
|
19
|
+
},
|
|
20
|
+
"daily_coding": {
|
|
21
|
+
"first": [
|
|
22
|
+
"gpt-6.1-sol",
|
|
23
|
+
"gemini-4-carbon",
|
|
24
|
+
"claude-sonnet-5"
|
|
25
|
+
],
|
|
26
|
+
"backup": [
|
|
27
|
+
"gemini-3.8-flash",
|
|
28
|
+
"claude-sonnet-4-6",
|
|
29
|
+
"deepseek-v4.1-flash"
|
|
30
|
+
]
|
|
31
|
+
},
|
|
32
|
+
"fast_cheap_bulk": {
|
|
33
|
+
"first": [
|
|
34
|
+
"deepseek-v4.1-flash",
|
|
35
|
+
"gemini-3.8-flash"
|
|
36
|
+
],
|
|
37
|
+
"backup": [
|
|
38
|
+
"gpt-6.1-terra",
|
|
39
|
+
"glm-5.3-flash",
|
|
40
|
+
"gpt-6-luna",
|
|
41
|
+
"gpt-oss-120b"
|
|
42
|
+
]
|
|
43
|
+
},
|
|
44
|
+
"review_find_bugs": {
|
|
45
|
+
"first": [
|
|
46
|
+
"gpt-6-astra",
|
|
47
|
+
"gemini-4-hydrogen",
|
|
48
|
+
"claude-opus-5"
|
|
49
|
+
],
|
|
50
|
+
"backup": [
|
|
51
|
+
"gemini-4-argon",
|
|
52
|
+
"claude-opus-4-6",
|
|
53
|
+
"gpt-6.1-sol",
|
|
54
|
+
"glm-5.3"
|
|
55
|
+
],
|
|
56
|
+
"note": "Prefer a different vendor from the writer."
|
|
57
|
+
},
|
|
58
|
+
"vision_ui_computer_use": {
|
|
59
|
+
"first": [
|
|
60
|
+
"gpt-6.1-terra",
|
|
61
|
+
"gemini-4-argon",
|
|
62
|
+
"gemini-4-carbon"
|
|
63
|
+
],
|
|
64
|
+
"backup": [
|
|
65
|
+
"gemini-3.8-flash"
|
|
66
|
+
]
|
|
67
|
+
},
|
|
68
|
+
"research": {
|
|
69
|
+
"first": [
|
|
70
|
+
"gemini-4-argon",
|
|
71
|
+
"gemini-4-hydrogen",
|
|
72
|
+
"gpt-6-astra"
|
|
73
|
+
],
|
|
74
|
+
"backup": [
|
|
75
|
+
"kimi-k3.1-pro",
|
|
76
|
+
"claude-opus-4-6"
|
|
77
|
+
]
|
|
78
|
+
},
|
|
79
|
+
"docs_simple_writing": {
|
|
80
|
+
"first": [
|
|
81
|
+
"gpt-6-luna",
|
|
82
|
+
"gemini-3.8-flash",
|
|
83
|
+
"deepseek-v4.1-flash"
|
|
84
|
+
],
|
|
85
|
+
"backup": [
|
|
86
|
+
"gpt-oss-120b",
|
|
87
|
+
"glm-5.3-flash",
|
|
88
|
+
"claude-haiku-5-5",
|
|
89
|
+
"claude-sonnet-4-6"
|
|
90
|
+
]
|
|
91
|
+
},
|
|
92
|
+
"chinese": {
|
|
93
|
+
"first": [
|
|
94
|
+
"glm-5.3",
|
|
95
|
+
"deepseek-v4.1-flash",
|
|
96
|
+
"qwen3.8-max"
|
|
97
|
+
],
|
|
98
|
+
"backup": [
|
|
99
|
+
"gemini-4-carbon",
|
|
100
|
+
"kimi-k3.1-pro"
|
|
101
|
+
]
|
|
102
|
+
}
|
|
103
|
+
},
|
|
104
|
+
"effort_advice": "Most developers default to medium. Raise to high/xhigh/max for hard bugs, architecture and long multi-step tasks; for routine edits, docs and bulk work high effort is slower and rarely worth it.",
|
|
105
|
+
"models": [
|
|
106
|
+
{
|
|
107
|
+
"id": "gpt-6-astra",
|
|
108
|
+
"vendor": "OpenAI",
|
|
109
|
+
"tier": "flagship",
|
|
110
|
+
"grade": "S",
|
|
111
|
+
"speed": "medium-slow",
|
|
112
|
+
"best_for": [
|
|
113
|
+
"hardest long tasks",
|
|
114
|
+
"design decisions",
|
|
115
|
+
"review",
|
|
116
|
+
"research"
|
|
117
|
+
],
|
|
118
|
+
"summary": {
|
|
119
|
+
"en": "OpenAI's strongest model for hard end-to-end agent work; some developers find it faster than Claude Opus on long tasks. Expensive; save for the hardest work.",
|
|
120
|
+
"zh": "OpenAI 最强,最难的端到端 agent 活;有开发者觉得长任务比 Opus 还快。贵,留给最难的活。"
|
|
121
|
+
},
|
|
122
|
+
"confidence": "confirmed",
|
|
123
|
+
"modalities": {
|
|
124
|
+
"input": [
|
|
125
|
+
"text",
|
|
126
|
+
"image"
|
|
127
|
+
],
|
|
128
|
+
"output": [
|
|
129
|
+
"text"
|
|
130
|
+
],
|
|
131
|
+
"context_tokens": 1050000,
|
|
132
|
+
"max_output_tokens": 128000
|
|
133
|
+
}
|
|
134
|
+
},
|
|
135
|
+
{
|
|
136
|
+
"id": "gpt-6.1-sol",
|
|
137
|
+
"vendor": "OpenAI",
|
|
138
|
+
"tier": "workhorse",
|
|
139
|
+
"grade": "A",
|
|
140
|
+
"speed": "medium",
|
|
141
|
+
"best_for": [
|
|
142
|
+
"writing code",
|
|
143
|
+
"large repos",
|
|
144
|
+
"long tasks that run tests until green",
|
|
145
|
+
"review"
|
|
146
|
+
],
|
|
147
|
+
"summary": {
|
|
148
|
+
"en": "Balanced main coder: close to Astra on agentic coding at a fraction of the cost. medium for routine edits, high/xhigh for hard bugs. Occasionally takes shortcuts at low effort.",
|
|
149
|
+
"zh": "平衡的主力写手:agentic 编程接近 Astra,成本低得多。日常 medium,难 bug 开 high/xhigh。低档偶尔偷懒。"
|
|
150
|
+
},
|
|
151
|
+
"confidence": "confirmed",
|
|
152
|
+
"modalities": {
|
|
153
|
+
"input": [
|
|
154
|
+
"text",
|
|
155
|
+
"image"
|
|
156
|
+
],
|
|
157
|
+
"output": [
|
|
158
|
+
"text"
|
|
159
|
+
],
|
|
160
|
+
"context_tokens": 1050000,
|
|
161
|
+
"max_output_tokens": 128000
|
|
162
|
+
}
|
|
163
|
+
},
|
|
164
|
+
{
|
|
165
|
+
"id": "gpt-6.1-terra",
|
|
166
|
+
"vendor": "OpenAI",
|
|
167
|
+
"tier": "fast",
|
|
168
|
+
"grade": "A",
|
|
169
|
+
"speed": "fast",
|
|
170
|
+
"best_for": [
|
|
171
|
+
"vision",
|
|
172
|
+
"computer use",
|
|
173
|
+
"everyday coding",
|
|
174
|
+
"docs"
|
|
175
|
+
],
|
|
176
|
+
"summary": {
|
|
177
|
+
"en": "Cheaper, faster sibling of Sol with strong vision and computer use. Good execution layer; use Sol or Astra for the hardest pure coding.",
|
|
178
|
+
"zh": "Sol 的便宜快速版,看图和电脑操作强。适合当执行层;最难的纯编程用 Sol 或 Astra。"
|
|
179
|
+
},
|
|
180
|
+
"confidence": "confirmed",
|
|
181
|
+
"modalities": {
|
|
182
|
+
"input": [
|
|
183
|
+
"text",
|
|
184
|
+
"image"
|
|
185
|
+
],
|
|
186
|
+
"output": [
|
|
187
|
+
"text"
|
|
188
|
+
],
|
|
189
|
+
"context_tokens": 1050000,
|
|
190
|
+
"max_output_tokens": 128000
|
|
191
|
+
}
|
|
192
|
+
},
|
|
193
|
+
{
|
|
194
|
+
"id": "gpt-6-luna",
|
|
195
|
+
"vendor": "OpenAI",
|
|
196
|
+
"tier": "cheap",
|
|
197
|
+
"grade": "B",
|
|
198
|
+
"speed": "fast",
|
|
199
|
+
"best_for": [
|
|
200
|
+
"simple docs",
|
|
201
|
+
"extraction",
|
|
202
|
+
"high-volume chores"
|
|
203
|
+
],
|
|
204
|
+
"summary": {
|
|
205
|
+
"en": "Cheapest GPT-6 tier for simple, high-volume work. Weak on complex coding.",
|
|
206
|
+
"zh": "GPT-6 最便宜一档,简单量大的活。复杂编程弱。"
|
|
207
|
+
},
|
|
208
|
+
"confidence": "confirmed",
|
|
209
|
+
"modalities": {
|
|
210
|
+
"input": [
|
|
211
|
+
"text",
|
|
212
|
+
"image"
|
|
213
|
+
],
|
|
214
|
+
"output": [
|
|
215
|
+
"text"
|
|
216
|
+
],
|
|
217
|
+
"context_tokens": 1050000,
|
|
218
|
+
"max_output_tokens": 128000
|
|
219
|
+
}
|
|
220
|
+
},
|
|
221
|
+
{
|
|
222
|
+
"id": "gpt-6-sol",
|
|
223
|
+
"vendor": "OpenAI",
|
|
224
|
+
"tier": "superseded",
|
|
225
|
+
"grade": "A-",
|
|
226
|
+
"speed": "medium",
|
|
227
|
+
"summary": {
|
|
228
|
+
"en": "Superseded by gpt-6.1-sol; prefer that.",
|
|
229
|
+
"zh": "已被 gpt-6.1-sol 取代,优先用它。"
|
|
230
|
+
},
|
|
231
|
+
"confidence": "confirmed"
|
|
232
|
+
},
|
|
233
|
+
{
|
|
234
|
+
"id": "gpt-5.6-sol",
|
|
235
|
+
"aliases": [
|
|
236
|
+
"gpt-5.6-terra",
|
|
237
|
+
"gpt-5.6-luna"
|
|
238
|
+
],
|
|
239
|
+
"vendor": "OpenAI",
|
|
240
|
+
"tier": "previous generation",
|
|
241
|
+
"grade": "B",
|
|
242
|
+
"speed": "medium",
|
|
243
|
+
"summary": {
|
|
244
|
+
"en": "Previous generation (July 2026), clearly behind GPT-6.x. Use only if newer models are unavailable.",
|
|
245
|
+
"zh": "上一代(2026-07),明显落后 GPT-6.x,没有新模型时才用。"
|
|
246
|
+
},
|
|
247
|
+
"confidence": "confirmed"
|
|
248
|
+
},
|
|
249
|
+
{
|
|
250
|
+
"id": "gemini-4-argon",
|
|
251
|
+
"vendor": "Google",
|
|
252
|
+
"tier": "flagship",
|
|
253
|
+
"grade": "S",
|
|
254
|
+
"speed": "medium",
|
|
255
|
+
"best_for": [
|
|
256
|
+
"hardest coding",
|
|
257
|
+
"research",
|
|
258
|
+
"long context",
|
|
259
|
+
"multimodal"
|
|
260
|
+
],
|
|
261
|
+
"summary": {
|
|
262
|
+
"en": "Google's flagship; top-tier on agentic coding benchmarks and strong at long context and multimodal. Less discussed for daily coding than Claude or GPT.",
|
|
263
|
+
"zh": "Google 旗舰,agentic 编程榜单顶尖,长上下文和多模态强。日常写代码讨论少于 Claude 和 GPT。"
|
|
264
|
+
},
|
|
265
|
+
"confidence": "confirmed",
|
|
266
|
+
"modalities": {
|
|
267
|
+
"input": [
|
|
268
|
+
"text",
|
|
269
|
+
"image",
|
|
270
|
+
"audio",
|
|
271
|
+
"video",
|
|
272
|
+
"pdf"
|
|
273
|
+
],
|
|
274
|
+
"output": [
|
|
275
|
+
"text"
|
|
276
|
+
],
|
|
277
|
+
"context_tokens": 1048576,
|
|
278
|
+
"max_output_tokens": 65536
|
|
279
|
+
}
|
|
280
|
+
},
|
|
281
|
+
{
|
|
282
|
+
"id": "gemini-3.8-flash",
|
|
283
|
+
"aliases": [
|
|
284
|
+
"gemini-3.7-flash"
|
|
285
|
+
],
|
|
286
|
+
"vendor": "Google",
|
|
287
|
+
"tier": "fast",
|
|
288
|
+
"grade": "A",
|
|
289
|
+
"speed": "fast",
|
|
290
|
+
"best_for": [
|
|
291
|
+
"writing code",
|
|
292
|
+
"long agent tasks",
|
|
293
|
+
"vision",
|
|
294
|
+
"good value"
|
|
295
|
+
],
|
|
296
|
+
"summary": {
|
|
297
|
+
"en": "Fast, good-value coder for long agent tasks and vision; slightly behind Argon on the hardest reasoning. 3.7 is a bit weaker.",
|
|
298
|
+
"zh": "又快又划算,长 agent 任务和看图都行;最难推理略逊 Argon。3.7 稍弱。"
|
|
299
|
+
},
|
|
300
|
+
"confidence": "confirmed",
|
|
301
|
+
"modalities": {
|
|
302
|
+
"input": [
|
|
303
|
+
"text",
|
|
304
|
+
"image",
|
|
305
|
+
"audio",
|
|
306
|
+
"video",
|
|
307
|
+
"pdf"
|
|
308
|
+
],
|
|
309
|
+
"output": [
|
|
310
|
+
"text"
|
|
311
|
+
],
|
|
312
|
+
"context_tokens": 1048576,
|
|
313
|
+
"max_output_tokens": 65536
|
|
314
|
+
}
|
|
315
|
+
},
|
|
316
|
+
{
|
|
317
|
+
"id": "grok-4.7",
|
|
318
|
+
"vendor": "xAI",
|
|
319
|
+
"tier": "flagship",
|
|
320
|
+
"grade": "A",
|
|
321
|
+
"speed": "medium (fast variants exist)",
|
|
322
|
+
"best_for": [
|
|
323
|
+
"writing code",
|
|
324
|
+
"long multi-hour tasks",
|
|
325
|
+
"review"
|
|
326
|
+
],
|
|
327
|
+
"summary": {
|
|
328
|
+
"en": "xAI's flagship, clearly better than 4.6 at long agent runs; some boards still trail the top Claude and GPT models. Fast variants trade depth for speed.",
|
|
329
|
+
"zh": "xAI 旗舰,长时间 agent 明显强于 4.6;部分榜仍落后最强的 Claude 和 GPT。fast 版拿深度换速度。"
|
|
330
|
+
},
|
|
331
|
+
"confidence": "confirmed",
|
|
332
|
+
"modalities": {
|
|
333
|
+
"input": [
|
|
334
|
+
"text",
|
|
335
|
+
"image"
|
|
336
|
+
],
|
|
337
|
+
"output": [
|
|
338
|
+
"text"
|
|
339
|
+
],
|
|
340
|
+
"context_tokens": 500000,
|
|
341
|
+
"max_output_tokens": null
|
|
342
|
+
}
|
|
343
|
+
},
|
|
344
|
+
{
|
|
345
|
+
"id": "composer-2.5",
|
|
346
|
+
"vendor": "Cursor",
|
|
347
|
+
"tier": "specialized",
|
|
348
|
+
"grade": "A",
|
|
349
|
+
"speed": "fast",
|
|
350
|
+
"best_for": [
|
|
351
|
+
"coding inside cursor-agent",
|
|
352
|
+
"large repos"
|
|
353
|
+
],
|
|
354
|
+
"summary": {
|
|
355
|
+
"en": "Cursor's own model, tuned for its tools and terminal; good value inside cursor-agent, not used elsewhere.",
|
|
356
|
+
"zh": "Cursor 自家模型,针对它的工具和终端调过;在 cursor-agent 里划算,别处用不到。"
|
|
357
|
+
},
|
|
358
|
+
"confidence": "confirmed",
|
|
359
|
+
"modalities": {
|
|
360
|
+
"input": [
|
|
361
|
+
"text"
|
|
362
|
+
],
|
|
363
|
+
"output": [
|
|
364
|
+
"text"
|
|
365
|
+
],
|
|
366
|
+
"context_tokens": 200000,
|
|
367
|
+
"max_output_tokens": null
|
|
368
|
+
}
|
|
369
|
+
},
|
|
370
|
+
{
|
|
371
|
+
"id": "deepseek-v4.1-flash",
|
|
372
|
+
"aliases": [
|
|
373
|
+
"deepseek-flash"
|
|
374
|
+
],
|
|
375
|
+
"vendor": "DeepSeek",
|
|
376
|
+
"tier": "fast",
|
|
377
|
+
"grade": "A",
|
|
378
|
+
"speed": "fast",
|
|
379
|
+
"best_for": [
|
|
380
|
+
"execution layer",
|
|
381
|
+
"bulk coding",
|
|
382
|
+
"long sessions",
|
|
383
|
+
"Chinese"
|
|
384
|
+
],
|
|
385
|
+
"summary": {
|
|
386
|
+
"en": "Very popular low-cost executor: many developers let Claude plan and review while DeepSeek writes the code. 1M context. Can queue at peak hours and is shallower than Claude on hard problems.",
|
|
387
|
+
"zh": "很热门的便宜执行层:很多人让 Claude 规划和审查,DeepSeek 写代码。1M 上下文。高峰会排队,难题深度不如 Claude。"
|
|
388
|
+
},
|
|
389
|
+
"confidence": "confirmed",
|
|
390
|
+
"modalities": {
|
|
391
|
+
"input": [
|
|
392
|
+
"text",
|
|
393
|
+
"image"
|
|
394
|
+
],
|
|
395
|
+
"output": [
|
|
396
|
+
"text"
|
|
397
|
+
],
|
|
398
|
+
"context_tokens": 1000000,
|
|
399
|
+
"max_output_tokens": 384000
|
|
400
|
+
}
|
|
401
|
+
},
|
|
402
|
+
{
|
|
403
|
+
"id": "deepseek-v4-pro",
|
|
404
|
+
"vendor": "DeepSeek",
|
|
405
|
+
"tier": "being retired",
|
|
406
|
+
"grade": "A",
|
|
407
|
+
"speed": "medium",
|
|
408
|
+
"summary": {
|
|
409
|
+
"en": "Being replaced by deepseek-v4.1-flash, which beats it on several tasks; prefer the flash model.",
|
|
410
|
+
"zh": "正被 deepseek-v4.1-flash 取代,后者多项更强,优先用 flash。"
|
|
411
|
+
},
|
|
412
|
+
"confidence": "confirmed"
|
|
413
|
+
},
|
|
414
|
+
{
|
|
415
|
+
"id": "glm-5.3",
|
|
416
|
+
"vendor": "Zhipu",
|
|
417
|
+
"tier": "workhorse",
|
|
418
|
+
"grade": "A",
|
|
419
|
+
"speed": "medium",
|
|
420
|
+
"best_for": [
|
|
421
|
+
"writing code",
|
|
422
|
+
"review",
|
|
423
|
+
"Chinese"
|
|
424
|
+
],
|
|
425
|
+
"summary": {
|
|
426
|
+
"en": "Strong open-weight coder, widely used in the Chinese community; still behind the top US models on the hardest tasks.",
|
|
427
|
+
"zh": "很强的开源写手,中文圈用得多;最难的活仍落后美国顶尖模型。"
|
|
428
|
+
},
|
|
429
|
+
"confidence": "confirmed",
|
|
430
|
+
"modalities": {
|
|
431
|
+
"input": [
|
|
432
|
+
"text"
|
|
433
|
+
],
|
|
434
|
+
"output": [
|
|
435
|
+
"text"
|
|
436
|
+
],
|
|
437
|
+
"context_tokens": 1048576,
|
|
438
|
+
"max_output_tokens": 128000
|
|
439
|
+
}
|
|
440
|
+
},
|
|
441
|
+
{
|
|
442
|
+
"id": "glm-5.3-flash",
|
|
443
|
+
"aliases": [
|
|
444
|
+
"glm-5.3-highspeed"
|
|
445
|
+
],
|
|
446
|
+
"vendor": "Zhipu",
|
|
447
|
+
"tier": "fast",
|
|
448
|
+
"grade": "A-",
|
|
449
|
+
"speed": "fast",
|
|
450
|
+
"best_for": [
|
|
451
|
+
"cheap coding",
|
|
452
|
+
"vision",
|
|
453
|
+
"bulk",
|
|
454
|
+
"Chinese"
|
|
455
|
+
],
|
|
456
|
+
"summary": {
|
|
457
|
+
"en": "Cheap multimodal GLM for bulk and simple coding. highspeed is a speed-focused variant with little public feedback.",
|
|
458
|
+
"zh": "便宜的多模态 GLM,适合批量和简单编程。highspeed 是提速版,公开反馈很少。"
|
|
459
|
+
},
|
|
460
|
+
"confidence": "probable",
|
|
461
|
+
"modalities": {
|
|
462
|
+
"input": [
|
|
463
|
+
"text",
|
|
464
|
+
"image",
|
|
465
|
+
"video"
|
|
466
|
+
],
|
|
467
|
+
"output": [
|
|
468
|
+
"text"
|
|
469
|
+
],
|
|
470
|
+
"context_tokens": 1000000,
|
|
471
|
+
"max_output_tokens": 128000
|
|
472
|
+
}
|
|
473
|
+
},
|
|
474
|
+
{
|
|
475
|
+
"id": "kimi-k3.1-pro",
|
|
476
|
+
"aliases": [
|
|
477
|
+
"kimi-k3"
|
|
478
|
+
],
|
|
479
|
+
"vendor": "Moonshot",
|
|
480
|
+
"tier": "flagship",
|
|
481
|
+
"grade": "A",
|
|
482
|
+
"speed": "slow",
|
|
483
|
+
"best_for": [
|
|
484
|
+
"long tasks",
|
|
485
|
+
"long context",
|
|
486
|
+
"research"
|
|
487
|
+
],
|
|
488
|
+
"summary": {
|
|
489
|
+
"en": "Strong at long tasks and long context, near the frontier on terminal benchmarks; slow. Some users note a borrowed style. K3 is the weaker predecessor.",
|
|
490
|
+
"zh": "长任务和长上下文强,终端类榜单接近前沿;慢。有人吐槽风格像别家。K3 是较弱的前代。"
|
|
491
|
+
},
|
|
492
|
+
"confidence": "probable",
|
|
493
|
+
"modalities": {
|
|
494
|
+
"input": [
|
|
495
|
+
"text",
|
|
496
|
+
"image"
|
|
497
|
+
],
|
|
498
|
+
"output": [
|
|
499
|
+
"text"
|
|
500
|
+
],
|
|
501
|
+
"context_tokens": 1048576,
|
|
502
|
+
"max_output_tokens": null
|
|
503
|
+
}
|
|
504
|
+
},
|
|
505
|
+
{
|
|
506
|
+
"id": "qwen3.8-max",
|
|
507
|
+
"vendor": "Alibaba",
|
|
508
|
+
"tier": "flagship",
|
|
509
|
+
"grade": "A",
|
|
510
|
+
"speed": "medium",
|
|
511
|
+
"best_for": [
|
|
512
|
+
"coding",
|
|
513
|
+
"Chinese",
|
|
514
|
+
"multimodal"
|
|
515
|
+
],
|
|
516
|
+
"summary": {
|
|
517
|
+
"en": "Alibaba's flagship; capable at coding, Chinese and multimodal, but rarely discussed as a daily coding model.",
|
|
518
|
+
"zh": "阿里旗舰,编程、中文、多模态都行,但很少有人拿它当日常写代码的主力。"
|
|
519
|
+
},
|
|
520
|
+
"confidence": "probable",
|
|
521
|
+
"modalities": {
|
|
522
|
+
"input": [
|
|
523
|
+
"text",
|
|
524
|
+
"image",
|
|
525
|
+
"video"
|
|
526
|
+
],
|
|
527
|
+
"output": [
|
|
528
|
+
"text"
|
|
529
|
+
],
|
|
530
|
+
"context_tokens": 1000000,
|
|
531
|
+
"max_output_tokens": 131072
|
|
532
|
+
}
|
|
533
|
+
},
|
|
534
|
+
{
|
|
535
|
+
"id": "claude-opus-5",
|
|
536
|
+
"aliases": [
|
|
537
|
+
"claude-opus-5-5",
|
|
538
|
+
"opus"
|
|
539
|
+
],
|
|
540
|
+
"vendor": "Anthropic",
|
|
541
|
+
"tier": "flagship",
|
|
542
|
+
"grade": "S",
|
|
543
|
+
"speed": "medium",
|
|
544
|
+
"best_for": [
|
|
545
|
+
"planning",
|
|
546
|
+
"architecture",
|
|
547
|
+
"large refactors",
|
|
548
|
+
"review"
|
|
549
|
+
],
|
|
550
|
+
"summary": {
|
|
551
|
+
"en": "Most trusted for complex refactors, planning and cross-vendor review; expensive. Available directly via Claude Code or third-party CLIs.",
|
|
552
|
+
"zh": "复杂重构、架构规划和跨厂审查最受信任,但贵。可直接经 Claude Code 或第三方 CLI 调用。"
|
|
553
|
+
},
|
|
554
|
+
"confidence": "confirmed",
|
|
555
|
+
"modalities": {
|
|
556
|
+
"input": [
|
|
557
|
+
"text",
|
|
558
|
+
"image",
|
|
559
|
+
"pdf"
|
|
560
|
+
],
|
|
561
|
+
"output": [
|
|
562
|
+
"text"
|
|
563
|
+
],
|
|
564
|
+
"context_tokens": 1000000,
|
|
565
|
+
"max_output_tokens": 128000
|
|
566
|
+
}
|
|
567
|
+
},
|
|
568
|
+
{
|
|
569
|
+
"id": "claude-sonnet-5",
|
|
570
|
+
"aliases": [
|
|
571
|
+
"claude-sonnet-5-5",
|
|
572
|
+
"sonnet"
|
|
573
|
+
],
|
|
574
|
+
"vendor": "Anthropic",
|
|
575
|
+
"tier": "workhorse",
|
|
576
|
+
"grade": "A",
|
|
577
|
+
"speed": "fast",
|
|
578
|
+
"summary": {
|
|
579
|
+
"en": "Fast, solid daily coder with good value.",
|
|
580
|
+
"zh": "又快又稳的日常写手,性价比好。"
|
|
581
|
+
},
|
|
582
|
+
"confidence": "confirmed",
|
|
583
|
+
"modalities": {
|
|
584
|
+
"input": [
|
|
585
|
+
"text",
|
|
586
|
+
"image",
|
|
587
|
+
"pdf"
|
|
588
|
+
],
|
|
589
|
+
"output": [
|
|
590
|
+
"text"
|
|
591
|
+
],
|
|
592
|
+
"context_tokens": 1000000,
|
|
593
|
+
"max_output_tokens": 128000
|
|
594
|
+
}
|
|
595
|
+
},
|
|
596
|
+
{
|
|
597
|
+
"id": "claude-fable-5",
|
|
598
|
+
"aliases": [
|
|
599
|
+
"claude-fable-5-1",
|
|
600
|
+
"fable"
|
|
601
|
+
],
|
|
602
|
+
"vendor": "Anthropic",
|
|
603
|
+
"tier": "flagship",
|
|
604
|
+
"grade": "S",
|
|
605
|
+
"speed": "medium-slow",
|
|
606
|
+
"summary": {
|
|
607
|
+
"en": "Anthropic's strongest and most expensive model, for the hardest reasoning and agent work.",
|
|
608
|
+
"zh": "Anthropic 最强也最贵,最难的推理和 agent 活。"
|
|
609
|
+
},
|
|
610
|
+
"confidence": "confirmed",
|
|
611
|
+
"modalities": {
|
|
612
|
+
"input": [
|
|
613
|
+
"text",
|
|
614
|
+
"image",
|
|
615
|
+
"pdf"
|
|
616
|
+
],
|
|
617
|
+
"output": [
|
|
618
|
+
"text"
|
|
619
|
+
],
|
|
620
|
+
"context_tokens": 1000000,
|
|
621
|
+
"max_output_tokens": 128000
|
|
622
|
+
}
|
|
623
|
+
},
|
|
624
|
+
{
|
|
625
|
+
"id": "claude-haiku-5-5",
|
|
626
|
+
"aliases": [
|
|
627
|
+
"haiku",
|
|
628
|
+
"claude-haiku-5"
|
|
629
|
+
],
|
|
630
|
+
"vendor": "Anthropic",
|
|
631
|
+
"tier": "cheap",
|
|
632
|
+
"grade": "B",
|
|
633
|
+
"speed": "fastest",
|
|
634
|
+
"summary": {
|
|
635
|
+
"en": "Fastest, cheapest Claude for simple high-volume work.",
|
|
636
|
+
"zh": "最快最便宜的 Claude,简单量大的活。"
|
|
637
|
+
},
|
|
638
|
+
"confidence": "confirmed",
|
|
639
|
+
"modalities": {
|
|
640
|
+
"input": [
|
|
641
|
+
"text",
|
|
642
|
+
"image",
|
|
643
|
+
"pdf"
|
|
644
|
+
],
|
|
645
|
+
"output": [
|
|
646
|
+
"text"
|
|
647
|
+
],
|
|
648
|
+
"context_tokens": 1000000,
|
|
649
|
+
"max_output_tokens": 128000
|
|
650
|
+
}
|
|
651
|
+
},
|
|
652
|
+
{
|
|
653
|
+
"id": "gemini-4-carbon",
|
|
654
|
+
"vendor": "Google",
|
|
655
|
+
"tier": "workhorse",
|
|
656
|
+
"grade": "A",
|
|
657
|
+
"speed": "medium",
|
|
658
|
+
"best_for": [
|
|
659
|
+
"writing code",
|
|
660
|
+
"agentic workflows",
|
|
661
|
+
"long context",
|
|
662
|
+
"multimodal"
|
|
663
|
+
],
|
|
664
|
+
"summary": {
|
|
665
|
+
"en": "Google's Gemini 4 workhorse in Antigravity: strong agentic coding and multimodal comprehension at balanced speed.",
|
|
666
|
+
"zh": "Google 在 Antigravity 里的 Gemini 4 主力写手,agentic 编程和多模态理解稳,速度均衡。"
|
|
667
|
+
},
|
|
668
|
+
"confidence": "confirmed",
|
|
669
|
+
"modalities": {
|
|
670
|
+
"input": [
|
|
671
|
+
"text",
|
|
672
|
+
"image",
|
|
673
|
+
"audio",
|
|
674
|
+
"video",
|
|
675
|
+
"pdf"
|
|
676
|
+
],
|
|
677
|
+
"output": [
|
|
678
|
+
"text"
|
|
679
|
+
],
|
|
680
|
+
"context_tokens": 1048576,
|
|
681
|
+
"max_output_tokens": 65536
|
|
682
|
+
}
|
|
683
|
+
},
|
|
684
|
+
{
|
|
685
|
+
"id": "gemini-4-hydrogen",
|
|
686
|
+
"vendor": "Google",
|
|
687
|
+
"tier": "flagship",
|
|
688
|
+
"grade": "S",
|
|
689
|
+
"speed": "medium-slow",
|
|
690
|
+
"best_for": [
|
|
691
|
+
"deep reasoning",
|
|
692
|
+
"hardest coding",
|
|
693
|
+
"research",
|
|
694
|
+
"multimodal"
|
|
695
|
+
],
|
|
696
|
+
"summary": {
|
|
697
|
+
"en": "Google's Gemini 4 deep-reasoning model (supports up to xhigh effort) for hard coding, debugging and multimodal research.",
|
|
698
|
+
"zh": "Google Gemini 4 深度推理档(支持到 xhigh),适合高难编程、排障与多模态调研。"
|
|
699
|
+
},
|
|
700
|
+
"confidence": "confirmed",
|
|
701
|
+
"modalities": {
|
|
702
|
+
"input": [
|
|
703
|
+
"text",
|
|
704
|
+
"image",
|
|
705
|
+
"audio",
|
|
706
|
+
"video",
|
|
707
|
+
"pdf"
|
|
708
|
+
],
|
|
709
|
+
"output": [
|
|
710
|
+
"text"
|
|
711
|
+
],
|
|
712
|
+
"context_tokens": 1048576,
|
|
713
|
+
"max_output_tokens": 65536
|
|
714
|
+
}
|
|
715
|
+
},
|
|
716
|
+
{
|
|
717
|
+
"id": "gemini-pro-orchestrator",
|
|
718
|
+
"aliases": [
|
|
719
|
+
"gemini-pro-orchestrator-argon"
|
|
720
|
+
],
|
|
721
|
+
"vendor": "Google",
|
|
722
|
+
"tier": "specialized",
|
|
723
|
+
"grade": "A",
|
|
724
|
+
"speed": "medium",
|
|
725
|
+
"best_for": [
|
|
726
|
+
"multi-step orchestration",
|
|
727
|
+
"subagent coordination",
|
|
728
|
+
"planning"
|
|
729
|
+
],
|
|
730
|
+
"summary": {
|
|
731
|
+
"en": "Antigravity's orchestration model (Carbon and Argon variants) tuned for multi-step planning and subagent delegation.",
|
|
732
|
+
"zh": "Antigravity 编排模型(含 Carbon 与 Argon 变体),针对多步规划与子代理协同调优。"
|
|
733
|
+
},
|
|
734
|
+
"confidence": "confirmed",
|
|
735
|
+
"modalities": {
|
|
736
|
+
"input": [
|
|
737
|
+
"text",
|
|
738
|
+
"image",
|
|
739
|
+
"audio",
|
|
740
|
+
"video",
|
|
741
|
+
"pdf"
|
|
742
|
+
],
|
|
743
|
+
"output": [
|
|
744
|
+
"text"
|
|
745
|
+
],
|
|
746
|
+
"context_tokens": 1048576,
|
|
747
|
+
"max_output_tokens": 65536
|
|
748
|
+
}
|
|
749
|
+
},
|
|
750
|
+
{
|
|
751
|
+
"id": "claude-sonnet-4-6",
|
|
752
|
+
"vendor": "Anthropic",
|
|
753
|
+
"tier": "workhorse",
|
|
754
|
+
"grade": "A-",
|
|
755
|
+
"speed": "fast",
|
|
756
|
+
"best_for": [
|
|
757
|
+
"writing code",
|
|
758
|
+
"review",
|
|
759
|
+
"daily tasks"
|
|
760
|
+
],
|
|
761
|
+
"summary": {
|
|
762
|
+
"en": "Anthropic's Claude Sonnet 4.6 coder exposed via third-party CLI quotas; dependable for daily coding and cross-vendor review, behind Sonnet 5.",
|
|
763
|
+
"zh": "经第三方 CLI 额度提供的 Claude Sonnet 4.6 写手,日常改代码和异源审查稳妥,略逊 Sonnet 5。"
|
|
764
|
+
},
|
|
765
|
+
"confidence": "confirmed",
|
|
766
|
+
"modalities": {
|
|
767
|
+
"input": [
|
|
768
|
+
"text",
|
|
769
|
+
"image",
|
|
770
|
+
"pdf"
|
|
771
|
+
],
|
|
772
|
+
"output": [
|
|
773
|
+
"text"
|
|
774
|
+
],
|
|
775
|
+
"context_tokens": 200000,
|
|
776
|
+
"max_output_tokens": 64000
|
|
777
|
+
}
|
|
778
|
+
},
|
|
779
|
+
{
|
|
780
|
+
"id": "claude-opus-4-6",
|
|
781
|
+
"aliases": [
|
|
782
|
+
"claude-opus-4-6-thinking"
|
|
783
|
+
],
|
|
784
|
+
"vendor": "Anthropic",
|
|
785
|
+
"tier": "workhorse",
|
|
786
|
+
"grade": "A",
|
|
787
|
+
"speed": "medium",
|
|
788
|
+
"best_for": [
|
|
789
|
+
"review",
|
|
790
|
+
"architecture",
|
|
791
|
+
"complex debugging"
|
|
792
|
+
],
|
|
793
|
+
"summary": {
|
|
794
|
+
"en": "Anthropic's Claude Opus 4.6 (including thinking variant) via third-party CLI quotas; strong independent reviewer when Claude 5 is unavailable.",
|
|
795
|
+
"zh": "经第三方 CLI 额度提供的 Claude Opus 4.6(含 thinking 变体),没有 Claude 5 时适合做独立深度审查。"
|
|
796
|
+
},
|
|
797
|
+
"confidence": "confirmed",
|
|
798
|
+
"modalities": {
|
|
799
|
+
"input": [
|
|
800
|
+
"text",
|
|
801
|
+
"image",
|
|
802
|
+
"pdf"
|
|
803
|
+
],
|
|
804
|
+
"output": [
|
|
805
|
+
"text"
|
|
806
|
+
],
|
|
807
|
+
"context_tokens": 200000,
|
|
808
|
+
"max_output_tokens": 64000
|
|
809
|
+
}
|
|
810
|
+
},
|
|
811
|
+
{
|
|
812
|
+
"id": "gpt-oss-120b",
|
|
813
|
+
"vendor": "OpenAI",
|
|
814
|
+
"tier": "cheap",
|
|
815
|
+
"grade": "B",
|
|
816
|
+
"speed": "fast",
|
|
817
|
+
"best_for": [
|
|
818
|
+
"routine edits",
|
|
819
|
+
"text processing",
|
|
820
|
+
"low-cost chores"
|
|
821
|
+
],
|
|
822
|
+
"summary": {
|
|
823
|
+
"en": "OpenAI's 120B open-weight reasoning model; fast and light on quota for routine tasks, weaker than frontier models on complex repos.",
|
|
824
|
+
"zh": "OpenAI 120B 开源权重推理模型,日常小活轻快省额度,复杂仓库弱于前沿主力。"
|
|
825
|
+
},
|
|
826
|
+
"confidence": "confirmed",
|
|
827
|
+
"modalities": {
|
|
828
|
+
"input": [
|
|
829
|
+
"text"
|
|
830
|
+
],
|
|
831
|
+
"output": [
|
|
832
|
+
"text"
|
|
833
|
+
],
|
|
834
|
+
"context_tokens": 128000,
|
|
835
|
+
"max_output_tokens": 32768
|
|
836
|
+
}
|
|
837
|
+
},
|
|
838
|
+
{
|
|
839
|
+
"id": "glm-4.6v",
|
|
840
|
+
"vendor": "Zhipu",
|
|
841
|
+
"tier": "previous generation",
|
|
842
|
+
"grade": "B",
|
|
843
|
+
"speed": "fast",
|
|
844
|
+
"summary": {
|
|
845
|
+
"en": "Previous-generation Zhipu vision model; clearly behind GLM-5.3 and GLM-5.3-flash. Switch worker defaults to glm-5.3.",
|
|
846
|
+
"zh": "智谱上一代视觉模型,明显落后 GLM-5.3 与 GLM-5.3-flash,默认模型请改用 glm-5.3。"
|
|
847
|
+
},
|
|
848
|
+
"confidence": "confirmed",
|
|
849
|
+
"modalities": {
|
|
850
|
+
"input": [
|
|
851
|
+
"text",
|
|
852
|
+
"image"
|
|
853
|
+
],
|
|
854
|
+
"output": [
|
|
855
|
+
"text"
|
|
856
|
+
],
|
|
857
|
+
"context_tokens": 128000,
|
|
858
|
+
"max_output_tokens": 16384
|
|
859
|
+
}
|
|
860
|
+
}
|
|
861
|
+
],
|
|
862
|
+
"modalities_source": "Grok, official model docs only. These are model-level API limits; a CLI or subscription may expose less (e.g. pi reports 272K/500K context for gpt-6.x via the Codex subscription). CLI-reported values win.",
|
|
863
|
+
"modalities_as_of": "2026-11-18"
|
|
864
|
+
}
|