tokenusage-cli 0.1.0 → 0.2.1

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.
Files changed (131) hide show
  1. package/CHANGELOG.md +46 -19
  2. package/LICENSE +22 -22
  3. package/README.md +123 -94
  4. package/THIRD_PARTY_NOTICES.md +140 -140
  5. package/bin/tokenusage.js +69 -69
  6. package/package.json +59 -59
  7. package/src/cli.js +146 -134
  8. package/src/commands/device-login.js +205 -205
  9. package/src/commands/diagnostics.js +38 -38
  10. package/src/commands/doctor.js +110 -110
  11. package/src/commands/init.js +2484 -2436
  12. package/src/commands/pause.js +77 -0
  13. package/src/commands/sessions.js +29 -29
  14. package/src/commands/status.js +1758 -1747
  15. package/src/commands/sync.js +8048 -7934
  16. package/src/commands/uninstall.js +322 -312
  17. package/src/lib/account-usage-dedup.js +267 -267
  18. package/src/lib/antigravity-paths.js +48 -48
  19. package/src/lib/ark-agent-plan-limits.js +438 -438
  20. package/src/lib/ark-coding-plan-limits.js +537 -537
  21. package/src/lib/async-queue-reader.js +138 -138
  22. package/src/lib/atomic-json.js +27 -27
  23. package/src/lib/bearer-token.js +11 -11
  24. package/src/lib/categorizer-utils.js +232 -232
  25. package/src/lib/claude-categorizer.js +1343 -1343
  26. package/src/lib/claude-config.js +279 -279
  27. package/src/lib/claude-user-identity.js +56 -56
  28. package/src/lib/cli-ui.js +179 -179
  29. package/src/lib/cloud-account.js +541 -541
  30. package/src/lib/cloud-device-token.js +141 -141
  31. package/src/lib/cloud-sync-prefs.js +14 -14
  32. package/src/lib/cloud-usage-reset.js +115 -115
  33. package/src/lib/codex-config.js +595 -595
  34. package/src/lib/codex-context-breakdown.js +806 -806
  35. package/src/lib/codex-model-attribution.js +128 -128
  36. package/src/lib/codex-rollout-parser.js +1471 -1471
  37. package/src/lib/codex-service-tier.js +49 -49
  38. package/src/lib/codex-token-refresh.js +140 -140
  39. package/src/lib/codex-token-usage.js +195 -195
  40. package/src/lib/codex-usage-record.js +195 -195
  41. package/src/lib/collection-pause.js +77 -0
  42. package/src/lib/command-runner.js +313 -313
  43. package/src/lib/commandcode-limits.js +317 -317
  44. package/src/lib/context-health.js +124 -124
  45. package/src/lib/cursor-config.js +605 -605
  46. package/src/lib/cursor-store.js +1918 -1918
  47. package/src/lib/debug-flags.js +7 -7
  48. package/src/lib/devin-limits.js +249 -249
  49. package/src/lib/devin-usage.js +212 -212
  50. package/src/lib/diagnostics.js +383 -383
  51. package/src/lib/doctor.js +385 -385
  52. package/src/lib/dumate.js +95 -95
  53. package/src/lib/fs.js +455 -455
  54. package/src/lib/function-url.js +14 -14
  55. package/src/lib/gemini-config.js +291 -291
  56. package/src/lib/git-outcomes.js +243 -243
  57. package/src/lib/grok-context-breakdown.js +890 -890
  58. package/src/lib/grok-hook.js +360 -360
  59. package/src/lib/grok-limits.js +865 -865
  60. package/src/lib/grok-usage.js +98 -98
  61. package/src/lib/heatmap-compact.js +114 -114
  62. package/src/lib/ingest-sanitize.js +41 -41
  63. package/src/lib/init-flow.js +42 -42
  64. package/src/lib/install-resolver.js +170 -170
  65. package/src/lib/jsonl-lines.js +134 -134
  66. package/src/lib/kimi-model-repair.js +173 -173
  67. package/src/lib/kimi-profile.js +61 -61
  68. package/src/lib/kimi-session.js +216 -216
  69. package/src/lib/local-api.js +3588 -3588
  70. package/src/lib/machine-id.js +238 -238
  71. package/src/lib/multi-install-parser.js +113 -113
  72. package/src/lib/notify-command.js +38 -38
  73. package/src/lib/omp-hook.js +589 -589
  74. package/src/lib/openclaw-cli.js +104 -104
  75. package/src/lib/openclaw-hook.js +375 -375
  76. package/src/lib/openclaw-session-plugin.js +560 -560
  77. package/src/lib/opencode-config.js +121 -121
  78. package/src/lib/opencode-go-limits.js +807 -807
  79. package/src/lib/outcomes-engine.js +253 -253
  80. package/src/lib/passive-mode.js +219 -219
  81. package/src/lib/pet-packages.js +722 -722
  82. package/src/lib/pricing/curated-overrides.json +822 -822
  83. package/src/lib/pricing/index.js +379 -379
  84. package/src/lib/pricing/litellm-fetcher.js +172 -172
  85. package/src/lib/pricing/matcher.js +456 -456
  86. package/src/lib/pricing/seed-snapshot.json +1 -1
  87. package/src/lib/progress.js +76 -76
  88. package/src/lib/project-usage-purge.js +117 -117
  89. package/src/lib/prompt.js +24 -24
  90. package/src/lib/provider-status.js +104 -104
  91. package/src/lib/proxy-env.js +489 -489
  92. package/src/lib/proxy-settings.js +163 -163
  93. package/src/lib/qoder-cn-cli-limits.js +119 -119
  94. package/src/lib/qoder-limits.js +880 -880
  95. package/src/lib/quiet-warnings.js +26 -26
  96. package/src/lib/rollout.js +26005 -25966
  97. package/src/lib/runtime-config.js +9 -4
  98. package/src/lib/scan-roots.js +277 -277
  99. package/src/lib/scheduled-task.js +145 -0
  100. package/src/lib/sentry.js +231 -231
  101. package/src/lib/session-analytics.js +2972 -2972
  102. package/src/lib/session-performance.js +237 -237
  103. package/src/lib/session-upload.js +379 -379
  104. package/src/lib/skill-usage.js +267 -267
  105. package/src/lib/skills-manager.js +1591 -1591
  106. package/src/lib/source-metadata.js +48 -48
  107. package/src/lib/sqlite-reader.js +215 -215
  108. package/src/lib/star-cta.js +73 -73
  109. package/src/lib/subscription-manager.js +293 -293
  110. package/src/lib/subscriptions.js +508 -508
  111. package/src/lib/sync-report.js +138 -0
  112. package/src/lib/telemetry.js +193 -193
  113. package/src/lib/tracker-paths.js +16 -16
  114. package/src/lib/trae-cn-config.js +548 -548
  115. package/src/lib/trae-db.js +127 -127
  116. package/src/lib/trae-snapshot.js +230 -230
  117. package/src/lib/trae-sqlite.js +143 -143
  118. package/src/lib/trae-usage.js +121 -121
  119. package/src/lib/upload-queue.js +134 -134
  120. package/src/lib/upload-throttle.js +171 -171
  121. package/src/lib/usage-limits.js +4552 -4552
  122. package/src/lib/wrapped-aggregator.js +236 -236
  123. package/src/lib/wsl-probe.js +281 -281
  124. package/src/lib/zcode-limits.js +1172 -1172
  125. package/src/lib/zcode-reset-credits.js +66 -66
  126. package/test/audit-fixes.test.js +200 -0
  127. package/test/cli-smoke.test.js +53 -53
  128. package/test/collection-controls.test.js +210 -0
  129. package/test/ingest-sanitize.test.js +162 -162
  130. package/test/quiet-warnings.test.js +53 -53
  131. package/test/session-upload.test.js +427 -427
@@ -1,822 +1,822 @@
1
- {
2
- "_meta": {
3
- "note": "Curated price overrides. Always wins over LiteLLM. Two reasons to live here: (1) self-defined alias names that LiteLLM will never carry (kiro-*, hy3-*, composer-*, kimi-for-coding, free-tier OpenRouter routes); (2) prices we want to pin even if LiteLLM has the model (e.g. cache_write fields LiteLLM often omits). Units: USD per million tokens. Edit this file to override pricing without redeploying.",
4
- "units": "usd_per_million_tokens",
5
- "deepseek_v4_pricing_note": "Official DeepSeek V4 pricing verified on 2026-08-21. Peak hours are 01:00-04:00 and 06:00-10:00 UTC; all other hours are off-peak at half price. Exact entries store peak rates. computeRowCost applies the time-of-use discount from each row's UTC hour_start.",
6
- "iflytek_maas_pricing_note": "iFlytek MaaS inference prices supplied for AStudio on 2026-08-31. AStudio homepage: https://agent.xfyun.cn/. Official pricing source: https://maas.xfyun.cn/modelSquare. Source values are RMB per million tokens, converted at 7.2 RMB/USD, then rounded to two decimal places for billing. Models without a dedicated cache-hit price use the normal input rate; cache writes also use the normal input rate. iFlytek MaaS prices are source-scoped and do not inherit DeepSeek time-of-use discounts. Spark-X2.5 entries (spark-x2.5, spark-x2.5-4b, spark-x2.5-1.7b) were read from the model square on 2026-10-08; spark-x2.5 is listed at a limited-time 30% rate (0.96 / 3.6 / 0.14 RMB input / output / cache-hit, list price 3.2 / 12 / 0.48), and the two small models are free."
7
- },
8
- "exact": {
9
- "claude-opus-5-5": {
10
- "input": 4,
11
- "output": 20,
12
- "cache_read": 0.2,
13
- "cache_write": 5,
14
- "note": "Official Anthropic pricing verified 2026-10-11: https://platform.claude.com/docs/en/about-claude/pricing"
15
- },
16
- "claude-opus-5.5": {
17
- "input": 4,
18
- "output": 20,
19
- "cache_read": 0.2,
20
- "cache_write": 5,
21
- "note": "Official Anthropic pricing verified 2026-10-11: https://platform.claude.com/docs/en/about-claude/pricing"
22
- },
23
- "claude-sonnet-5-5": {
24
- "input": 2,
25
- "output": 10,
26
- "cache_read": 0.1,
27
- "cache_write": 2.5,
28
- "note": "Official Anthropic pricing verified 2026-10-11: https://platform.claude.com/docs/en/about-claude/pricing"
29
- },
30
- "claude-sonnet-5.5": {
31
- "input": 2,
32
- "output": 10,
33
- "cache_read": 0.1,
34
- "cache_write": 2.5,
35
- "note": "Official Anthropic pricing verified 2026-10-11: https://platform.claude.com/docs/en/about-claude/pricing"
36
- },
37
- "k3-256k": {
38
- "input": 3,
39
- "output": 15,
40
- "cache_read": 0.3,
41
- "note": "Bare alias emitted by Kimi Code (modelAlias \"kimi-code/k3\" -> \"k3\"). Same rates as kimi-k3."
42
- },
43
- "cline-gateway-models": {
44
- "input": 0,
45
- "output": 0,
46
- "cache_read": 0,
47
- "cache_write": 0,
48
- "note": "Cline's own gateway namespaces (`cline-free/*` free tier, `cline-pass/*` flat-rate Cline Pass). Cline's shipped model catalog declares pricing 0 for every id in both namespaces, so a turn there has no marginal per-token cost — same convention as PI_SUBSCRIPTION_SOURCES. The fuzzy entries at the top of `fuzzy` route both prefixes here BEFORE the model names can match a public rate (`cline-pass/glm-5.3` must not bill GLM-5.3 list price, `cline-free/deepseek-v4.1-flash` must not bill DeepSeek). Verified against Cline 0.0.32 (/Applications/Cline.app model catalog + `usageDelta` cost, which stays 0 for both namespaces). A Cline turn that DOES report a positive per-call cost is honoured first because `cline` is listed in SOURCES_WITH_AUTHORITATIVE_COST in pricing/index.js. Remove or re-point these entries when Cline publishes per-token pricing for either namespace."
49
- },
50
- "claude-fable-5": {
51
- "input": 10,
52
- "output": 50,
53
- "cache_read": 1,
54
- "cache_write": 12.5,
55
- "note": "Fable 5 \u2014 Anthropic's top tier above Opus. Not yet in LiteLLM. $10/$50 per MTok; cache_read 0.1x, cache_write 1.25x. Remove once LiteLLM carries it."
56
- },
57
- "claude-opus-5": {
58
- "input": 5,
59
- "output": 25,
60
- "cache_read": 0.5,
61
- "cache_write": 6.25,
62
- "note": "Opus 5 (public 2026-07). Standard Opus tier, unchanged from 4.6/4.7/4.8: $5/$25 per MTok, cache_read 0.1x, cache_write 1.25x. Not yet in LiteLLM. Fast mode is billed at $10/$50 \u2014 see claude-opus-5-fast. Remove once LiteLLM carries it."
63
- },
64
- "claude-opus-5-fast": {
65
- "input": 10,
66
- "output": 50,
67
- "cache_read": 1,
68
- "cache_write": 12.5,
69
- "note": "Opus 5 fast mode (speed=\"fast\", beta fast-mode-2026-02-01) is priced at $10/$50 per MTok \u2014 2x standard Opus. Kept separate so a -fast model id is not silently priced at the standard tier. See claude-opus-5."
70
- },
71
- "claude-opus-4-8": {
72
- "input": 5,
73
- "output": 25,
74
- "cache_read": 0.5,
75
- "cache_write": 6.25,
76
- "note": "Opus 4.8 not yet in LiteLLM. Pin to the standard Opus tier (same as 4.6/4.7); remove once LiteLLM carries it."
77
- },
78
- "claude-sonnet-5": {
79
- "input": 2,
80
- "output": 10,
81
- "cache_read": 0.2,
82
- "cache_write": 2.5,
83
- "note": "Sonnet 5. LiteLLM added this model on ~2026-06-30 with matching rates, but pin explicitly so pricing is correct immediately rather than depending on the 24h local pricing-cache refresh cycle."
84
- },
85
- "claude-haiku-4-5": {
86
- "input": 1,
87
- "output": 5,
88
- "cache_read": 0.1,
89
- "cache_write": 1.25,
90
- "note": "Canonical undated alias emitted by Cursor. Pin to the same rates as claude-haiku-4-5-20251001 so local cost does not depend on which dated LiteLLM aliases are present in the current cache."
91
- },
92
- "gpt-6-astra": {
93
- "input": 10,
94
- "output": 50,
95
- "cache_read": 1,
96
- "cache_write": 12.5,
97
- "priority_multiplier": 2,
98
- "note": "Official Standard USD/MTok rates verified 2026-09-07: https://developers.openai.com/api/docs/models/gpt-6-astra. Requests above 272K input tokens use 2x input/cache and 1.5x output rates; computeRowCost applies this only to observed long-context request subsets. priority_multiplier is the Fast tier, whose published table (20/2/25/100 short, 40/4/50/150 long, verified 2026-09-12) is exactly 2x the Standard rate of the same context tier — it multiplies, it does not replace, the long-context premium. The 2.5x figure in the Codex speed docs is a subscription-credit rate and must not be used here. computeRowCost applies it only to observed priority request subsets; usage whose service tier the logs never recorded stays Standard. Cloud aggregates use Standard short-context estimates because they do not retain per-request context or service tier. Reasoning-effort suffixes use the same base rates."
99
- },
100
- "gpt-6.1-sol": {
101
- "input": 2,
102
- "output": 10,
103
- "cache_read": 0.1,
104
- "cache_write": 2.5,
105
- "priority_multiplier": 2,
106
- "note": "Official Standard USD/MTok rates verified 2026-10-02 (issue #737): https://developers.openai.com/api/docs/models/gpt-6.1-sol. Requests above 272K input use 2x input/cache and 1.5x output rates; Fast mode is 2x the applicable rates. Premiums apply only to observed request subsets. Cloud aggregates remain Standard short-context estimates."
107
- },
108
- "gpt-6-sol": {
109
- "input": 2,
110
- "output": 10,
111
- "cache_read": 0.2,
112
- "cache_write": 2.5,
113
- "priority_multiplier": 2,
114
- "note": "Official Standard USD/MTok rates verified 2026-09-24: https://developers.openai.com/api/docs/models/gpt-6-sol. Same rules as gpt-6-astra: requests above 272K input tokens use 2x input/cache and 1.5x output for the full request, applied only to observed long-context request subsets; Fast mode is published as 2x the applicable rates (priority_multiplier). Reasoning-effort suffixes use the same base rates."
115
- },
116
- "gpt-5.6-sol": {
117
- "input": 4,
118
- "output": 20,
119
- "cache_read": 0.4,
120
- "cache_write": 5,
121
- "note": "Official GPT-5.6 Sol standard pricing: $4 input, $0.40 cached input, $5 cache write, and $20 output per MTok. Requests above 272K input are repriced from their preserved per-request long-context subset. Codex emits gpt-5.6-sol plus reasoning-effort variants caught by fuzzy matching."
122
- },
123
- "gpt-5.6-terra": {
124
- "input": 2,
125
- "output": 12,
126
- "cache_read": 0.2,
127
- "cache_write": 2.5,
128
- "note": "GPT-5.6 balanced default tier. Official short-context rates: $2/$12 per MTok (not the earlier $2.5/$15 snapshot). See gpt-5.6-sol."
129
- },
130
- "gpt-5.6-luna": {
131
- "input": 0.2,
132
- "output": 1.2,
133
- "cache_read": 0.02,
134
- "cache_write": 0.25,
135
- "note": "GPT-5.6 lightweight tier. Official short-context rates: $0.20/$1.20 per MTok (previous curated values $1/$6 overcounted 5x). See gpt-5.6-sol."
136
- },
137
- "kiro-agent": {
138
- "input": 3,
139
- "output": 15,
140
- "cache_read": 0.3,
141
- "cache_write": 3.75
142
- },
143
- "kiro-cli-agent": {
144
- "input": 3,
145
- "output": 15,
146
- "cache_read": 0.3,
147
- "cache_write": 3.75
148
- },
149
- "hy3-preview-agent": {
150
- "input": 0.167,
151
- "output": 0.556,
152
- "cache_read": 0.056,
153
- "cache_write": 0.167,
154
- "note": "Tencent Hunyuan Hy3 preview (CodeBuddy/WorkBuddy backend). Official TokenHub rate: 1.2 / 0.4 (cache hit) / 4.0 RMB per MTok in/read/out, converted at ~7.2 RMB/USD. DeepSeek-style cache: no write surcharge, so cache_write = input."
155
- },
156
- "hy3-preview": {
157
- "input": 0.167,
158
- "output": 0.556,
159
- "cache_read": 0.056,
160
- "cache_write": 0.167,
161
- "note": "See hy3-preview-agent."
162
- },
163
- "qwen3.8-flash": {
164
- "input": 0.15,
165
- "output": 0.47,
166
- "cache_read": 0.016,
167
- "cache_write": 0.2,
168
- "note": "Alibaba Model Studio Singapore (International), USD per MTok, verified 2026-09-29: https://www.alibabacloud.com/help/en/model-studio/qwen3-8-flash . Reference API estimate; other regions/providers can have different rates. Issue #715."
169
- },
170
- "hy4-preview": {
171
- "input": 0.833,
172
- "output": 2.5,
173
- "cache_read": 0.042,
174
- "cache_write": 0.833,
175
- "note": "Tencent Hunyuan Hy4 preview (released 2026-08-28; WorkBuddy/CodeBuddy backend). Official TokenHub rate: 6 / 0.3 (cache hit) / 18 RMB per MTok in/read/out, converted at ~7.2 RMB/USD like hy3-preview. No published cache-write price, so cache_write = input. Issue #633."
176
- },
177
- "hy4-preview-agent": {
178
- "input": 0.833,
179
- "output": 2.5,
180
- "cache_read": 0.042,
181
- "cache_write": 0.833,
182
- "note": "See hy4-preview. CodeBuddy's auto-router emits the -agent suffixed form."
183
- },
184
- "composer-1": {
185
- "input": 1.25,
186
- "output": 10,
187
- "cache_read": 0.125
188
- },
189
- "composer-1.5": {
190
- "input": 3.5,
191
- "output": 17.5,
192
- "cache_read": 0.35
193
- },
194
- "composer-2": {
195
- "input": 0.5,
196
- "output": 2.5,
197
- "cache_read": 0.2
198
- },
199
- "composer-2-fast": {
200
- "input": 1.5,
201
- "output": 7.5,
202
- "cache_read": 0.15
203
- },
204
- "composer-2.5": {
205
- "input": 0.5,
206
- "output": 2.5,
207
- "cache_read": 0.2,
208
- "note": "Cursor Composer 2.5. Cursor Models & Pricing: $0.50/M input, $0.20/M cached input, $2.50/M output; no cache-write rate is published. https://cursor.com/docs/models-and-pricing"
209
- },
210
- "composer-2.5-fast": {
211
- "input": 3.0,
212
- "output": 15.0,
213
- "cache_read": 0.5,
214
- "note": "Cursor Composer 2.5 Fast. Cursor Models & Pricing: $3/M input, $0.50/M cached input, $15/M output. Kept separate from the cheaper base SKU."
215
- },
216
- "gemini-4-argon": {
217
- "input": 2.0,
218
- "output": 10.0,
219
- "cache_read": 0.1,
220
- "cache_write": 0,
221
- "note": "Google Gemini 4 Argon introductory list rate (through 2026-11-30) matching Cursor Other Models pool: $2/M input, $0.10/M cached input, $10/M output. Effort suffixes (-high/-medium/-low/-minimal) are stripped by normalizeCursorModel. Sources: https://cursor.com/docs/models-and-pricing and https://ai.google.dev/gemini-api/docs/pricing"
222
- },
223
- "MiniMax-M2.7": {
224
- "input": 0.3,
225
- "output": 1.2,
226
- "cache_read": 0.06,
227
- "cache_write": 0.375
228
- },
229
- "MiniMax-M2.7-highspeed": {
230
- "input": 0.6,
231
- "output": 2.4,
232
- "cache_read": 0.06,
233
- "cache_write": 0.375
234
- },
235
- "minimax-m3": {
236
- "input": 0.3,
237
- "output": 1.2,
238
- "cache_read": 0.06,
239
- "cache_write": 0
240
- },
241
- "deepseek-v4-flash": {
242
- "input": 0.44,
243
- "output": 1.32,
244
- "cache_read": 0.014,
245
- "cache_write": 0.44
246
- },
247
- "deepseek-v4-pro": {
248
- "input": 1.32,
249
- "output": 3.96,
250
- "cache_read": 0.044,
251
- "cache_write": 1.32
252
- },
253
- "deepseek-v4-flash-vision-exp": {
254
- "input": 0.44,
255
- "output": 1.32,
256
- "cache_read": 0.014,
257
- "cache_write": 0.44
258
- },
259
- "deepseek-v4.1-flash": {
260
- "input": 0.3,
261
- "output": 1.2,
262
- "cache_read": 0.006,
263
- "cache_write": 0.3,
264
- "note": "DeepSeek V4.1 Flash (official API id deepseek-flash, released 2026-09-10). deepseek-v4.1-flash is the id reported by OpenRouter, Command Code and WorkBuddy. Peak rates per https://api-docs.deepseek.com/quick_start/pricing/ ($0.30 / $1.20 / $0.006 cache read per MTok); computeRowCost halves them off-peak like the other DeepSeek V4 entries."
265
- },
266
- "deepseek-flash": {
267
- "input": 0.3,
268
- "output": 1.2,
269
- "cache_read": 0.006,
270
- "cache_write": 0.3,
271
- "note": "Official DeepSeek API id for DeepSeek-V4.1-Flash. Same rates as deepseek-v4.1-flash."
272
- },
273
- "deepseek-chat": {
274
- "input": 0.14,
275
- "output": 0.28,
276
- "cache_read": 0.0028,
277
- "cache_write": 0.14
278
- },
279
- "grok-build": {
280
- "input": 1.25,
281
- "output": 2.5,
282
- "cache_read": 0.2,
283
- "note": "Grok Build TUI fallback estimate for sessions without turn_completed.usage. Preferred path is turn_completed usage with per-model modelUsage splits."
284
- },
285
- "cursor-grok-4.5": {
286
- "input": 2.0,
287
- "output": 6.0,
288
- "cache_read": 0.5,
289
- "cache_write": 0,
290
- "note": "Cursor Grok 4.5 base rate. Cursor Models & Pricing: $2/M input, $0.50/M cached input, $6/M output; no cache-write rate is published."
291
- },
292
- "cursor-grok-4.5-fast": {
293
- "input": 4.0,
294
- "output": 18.0,
295
- "cache_read": 1.0,
296
- "cache_write": 0,
297
- "note": "Cursor Grok 4.5 Fast rate. Cursor publishes $4/M input and $18/M output; the Cursor usage breakdown reports $1/M cached input. Kept separate from the cheaper base SKU."
298
- },
299
- "cursor-grok-4.6": {
300
- "input": 2.0,
301
- "output": 6.0,
302
- "cache_read": 0.5,
303
- "cache_write": 0,
304
- "note": "Cursor Grok 4.6 base rate. Cursor Models & Pricing: $2/M input, $0.50/M cached input, $6/M output; no cache-write rate is published."
305
- },
306
- "cursor-grok-4.6-fast": {
307
- "input": 4.0,
308
- "output": 12.0,
309
- "cache_read": 1.0,
310
- "cache_write": 0,
311
- "note": "Cursor Grok 4.6 Fast rate. Cursor publishes $4/M input, $1/M cached input, $12/M output. Kept separate from the cheaper base SKU."
312
- },
313
- "cursor-grok-4.7": {
314
- "input": 2.0,
315
- "output": 6.0,
316
- "cache_read": 0.5,
317
- "cache_write": 0,
318
- "note": "Cursor Grok 4.7 base rate. Cursor Models & Pricing: $2/M input, $0.50/M cached input, $6/M output. Long-context (>256k) premiums are not applied here yet because Cursor CSV rows do not expose per-request long-context flags."
319
- },
320
- "cursor-grok-4.7-fast": {
321
- "input": 4.0,
322
- "output": 12.0,
323
- "cache_read": 1.0,
324
- "cache_write": 0,
325
- "note": "Cursor Grok 4.7 Fast rate. Cursor publishes $4/M input, $1/M cached input, $12/M output. Kept separate from the cheaper base SKU."
326
- },
327
- "grok-4-0709": {
328
- "input": 3.0,
329
- "output": 15.0,
330
- "cache_read": 0.75
331
- },
332
- "grok-4": {
333
- "input": 3.0,
334
- "output": 15.0,
335
- "cache_read": 0.75
336
- },
337
- "grok-4-latest": {
338
- "input": 3.0,
339
- "output": 15.0,
340
- "cache_read": 0.75
341
- },
342
- "grok-4-fast": {
343
- "input": 0.2,
344
- "output": 0.5,
345
- "cache_read": 0.05
346
- },
347
- "grok-4-fast-reasoning": {
348
- "input": 0.2,
349
- "output": 0.5,
350
- "cache_read": 0.05
351
- },
352
- "grok-4-fast-non-reasoning": {
353
- "input": 0.2,
354
- "output": 0.5,
355
- "cache_read": 0.05
356
- },
357
- "grok-4-1-fast-non-reasoning": {
358
- "input": 0.2,
359
- "output": 0.5,
360
- "cache_read": 0.05
361
- },
362
- "deepseek-reasoner": {
363
- "input": 0.14,
364
- "output": 0.28,
365
- "cache_read": 0.0028,
366
- "cache_write": 0.14
367
- },
368
- "kimi-for-coding": {
369
- "input": 0.6,
370
- "output": 2,
371
- "cache_read": 0.15
372
- },
373
- "kimi-k2.5": {
374
- "input": 0.6,
375
- "output": 2,
376
- "cache_read": 0.15
377
- },
378
- "kimi-k2.5-free": {
379
- "input": 0,
380
- "output": 0,
381
- "cache_read": 0
382
- },
383
- "kimi-k2.6": {
384
- "input": 0.95,
385
- "output": 4,
386
- "cache_read": 0.16
387
- },
388
- "kimi-k2.7-code": {
389
- "input": 0.95,
390
- "output": 4,
391
- "cache_read": 0.19
392
- },
393
- "kimi-k3": {
394
- "input": 3,
395
- "output": 15,
396
- "cache_read": 0.3,
397
- "note": "Kimi K3 (released 2026-07-16). Reported API rates: $3/M input, $15/M output, $0.30/M cached input (0.1x). Not yet in LiteLLM; remove once it carries k3. Kimi Code records the alias as bare \"k3\" (kimi-code/k3), hence the separate exact key below."
398
- },
399
- "k3": {
400
- "input": 3,
401
- "output": 15,
402
- "cache_read": 0.3,
403
- "note": "Bare alias emitted by Kimi Code (modelAlias \"kimi-code/k3\" -> \"k3\"). Same rates as kimi-k3."
404
- },
405
- "glm-5.3": {
406
- "input": 1.4,
407
- "output": 4.4,
408
- "cache_read": 0.26
409
- },
410
- "glm-5.3-flash": {
411
- "input": 0.15,
412
- "output": 0.5,
413
- "cache_read": 0.03
414
- },
415
- "glm-5.2": {
416
- "input": 1.4,
417
- "output": 4.4,
418
- "cache_read": 0.26
419
- },
420
- "glm-5.1": {
421
- "input": 1.4,
422
- "output": 4.4,
423
- "cache_read": 0.26
424
- },
425
- "glm-5": {
426
- "input": 1.0,
427
- "output": 3.2,
428
- "cache_read": 0.2
429
- },
430
- "glm-5-turbo": {
431
- "input": 1.2,
432
- "output": 4.0,
433
- "cache_read": 0.24
434
- },
435
- "glm-4.7": {
436
- "input": 0.6,
437
- "output": 2.2,
438
- "cache_read": 0.11
439
- },
440
- "glm-4.7-flashx": {
441
- "input": 0.07,
442
- "output": 0.4,
443
- "cache_read": 0.01
444
- },
445
- "glm-4.7-flash": {
446
- "input": 0,
447
- "output": 0,
448
- "cache_read": 0
449
- },
450
- "glm-4.6": {
451
- "input": 0.6,
452
- "output": 2.2,
453
- "cache_read": 0.11
454
- },
455
- "glm-4.5": {
456
- "input": 0.6,
457
- "output": 2.2,
458
- "cache_read": 0.11
459
- },
460
- "glm-4.5-x": {
461
- "input": 2.2,
462
- "output": 8.9,
463
- "cache_read": 0.45
464
- },
465
- "glm-4.5-air": {
466
- "input": 0.2,
467
- "output": 1.1,
468
- "cache_read": 0.03
469
- },
470
- "glm-4.5-airx": {
471
- "input": 1.1,
472
- "output": 4.5,
473
- "cache_read": 0.22
474
- },
475
- "glm-4.5-flash": {
476
- "input": 0,
477
- "output": 0,
478
- "cache_read": 0
479
- },
480
- "glm-4.7-free": {
481
- "input": 0,
482
- "output": 0,
483
- "cache_read": 0
484
- },
485
- "nemotron-3-super-free": {
486
- "input": 0,
487
- "output": 0,
488
- "cache_read": 0
489
- },
490
- "mimo-v2-pro-free": {
491
- "input": 0,
492
- "output": 0,
493
- "cache_read": 0
494
- },
495
- "minimax-m2.1-free": {
496
- "input": 0,
497
- "output": 0,
498
- "cache_read": 0
499
- },
500
- "MiniMax-M2.1": {
501
- "input": 0.5,
502
- "output": 3,
503
- "cache_read": 0.05
504
- },
505
- "antigravity-gpt-oss-120b": {
506
- "input": 2.5,
507
- "output": 10,
508
- "cache_read": 0,
509
- "note": "Antigravity bridge alias. Approximate with GPT-4o-class pricing until Google exposes route-specific billing metadata."
510
- },
511
- "sakana/fugu-ultra": {
512
- "input": 5,
513
- "output": 30,
514
- "cache_read": 0.5,
515
- "cache_write": 5,
516
- "note": "Sakana Fugu Ultra \u2014 multi-agent orchestration via an OpenAI-compatible API (sakana.ai PAYG + OpenRouter), used through Codex/Cursor/Cline/ZCode etc. OpenRouter rates: $5/$30 per MTok in/out, cache_read $0.5/M; no published cache-write surcharge, so cache_write = input. Subscriptions ($20/$100/$200) bill against this token-equivalent rate."
517
- },
518
- "longcat-2.0": {
519
- "input": 0.278,
520
- "output": 1.111,
521
- "cache_read": 0.00556,
522
- "cache_write": 0.278,
523
- "note": "Meituan LongCat-2.0, seen via ZCode custom-provider routing (issue #276). Official longcat.chat/platform launch-promo rate: RMB 2 / 0.04 (cache hit) / 8 per MTok in/read/out, converted at ~7.2 RMB/USD. Standard list price is RMB 5/0.10/20 (2.5x) once the promo ends \u2014 re-verify against longcat.chat/platform/docs/zh/Pricing/LongCat-2.0.html before assuming either is permanent. DeepSeek-style cache: no write surcharge, so cache_write = input."
524
- },
525
- "step-3.7-flash": {
526
- "input": 0.2,
527
- "output": 1.15,
528
- "cache_read": 0.04,
529
- "cache_write": 0.2,
530
- "note": "StepFun Step 3.7 Flash (issue #283). Official platform.stepfun.ai pricing: $0.20 / $0.04 (cache hit) / $1.15 per MTok in/read/out. No published cache-write surcharge, so cache_write = input."
531
- },
532
- "step-3.5-flash": {
533
- "input": 0.1,
534
- "output": 0.3,
535
- "cache_read": 0.02,
536
- "cache_write": 0.1,
537
- "note": "StepFun Step 3.5 Flash incl. dated snapshots like step-3.5-flash-2603 (issue #283). Official platform.stepfun.ai pricing: $0.10 / $0.02 (cache hit) / $0.30 per MTok in/read/out. No published cache-write surcharge, so cache_write = input."
538
- },
539
- "grok-4.5-build": {
540
- "input": 2.0,
541
- "output": 6.0,
542
- "cache_read": 0.5,
543
- "cache_write": 0,
544
- "note": "Grok Build paid SKU (grok-4.5-build from turn_completed.modelUsage). Rates aligned with xAI/Cursor Grok 4.5 until a dedicated Build price list is published."
545
- },
546
- "grok-build-free": {
547
- "input": 0,
548
- "output": 0,
549
- "cache_read": 0,
550
- "cache_write": 0,
551
- "note": "Grok Build free tier (canonical model id). $0 marginal cost. Prefer this over grok-4.5-build-free so pricing lookup cannot fuzzy-match paid grok-4.5."
552
- },
553
- "grok-4.5-build-free": {
554
- "input": 0,
555
- "output": 0,
556
- "cache_read": 0,
557
- "cache_write": 0,
558
- "note": "Alias for free Build SKU as labeled in Grok updates.jsonl modelUsage. Canonical storage id is grok-build-free."
559
- }
560
- },
561
- "alias": {
562
- "auto": "composer-1"
563
- },
564
- "source_exact": {
565
- "acode": {
566
- "spark-x2.5": { "input": 0.13, "output": 0.50, "cache_read": 0.02, "cache_write": 0.13 },
567
- "spark-x2.5-4b": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
568
- "spark-x2.5-1.7b": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
569
- "xopglm53": { "input": 1.11, "output": 3.89, "cache_read": 0.28, "cache_write": 1.11 },
570
- "xopdeepseekv4pro0813": { "input": 1.25, "output": 3.75, "cache_read": 0.04, "cache_write": 1.25 },
571
- "xopdeepseekv4flash0731": { "input": 0.14, "output": 0.28, "cache_read": 0.03, "cache_write": 0.14 },
572
- "xopkimik27code": { "input": 0.90, "output": 3.75, "cache_read": 0.90, "cache_write": 0.90 },
573
- "xopglm52": { "input": 1.11, "output": 3.89, "cache_read": 0.28, "cache_write": 1.11 },
574
- "xopdeepseekv4flash": { "input": 0.14, "output": 0.28, "cache_read": 0.03, "cache_write": 0.14 },
575
- "xopkimik26": { "input": 0.90, "output": 3.75, "cache_read": 0.18, "cache_write": 0.90 },
576
- "xopdeepseekv4pro": { "input": 1.67, "output": 3.33, "cache_read": 0.14, "cache_write": 1.67 },
577
- "xopqwen36v35b": { "input": 0.15, "output": 0.90, "cache_read": 0.15, "cache_write": 0.15 },
578
- "xophunyuan7bmt": { "input": 0.07, "output": 0.28, "cache_read": 0.07, "cache_write": 0.07 },
579
- "xoppaddleocrv16": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
580
- "xsparkx2flash": { "input": 0.14, "output": 0.28, "cache_read": 0.14, "cache_write": 0.14 },
581
- "xopglm51": { "input": 1.11, "output": 3.89, "cache_read": 0.22, "cache_write": 1.11 },
582
- "xsparkx2": { "input": 0.42, "output": 0.42, "cache_read": 0.42, "cache_write": 0.42 },
583
- "xop35qwen2b": { "input": 0.03, "output": 0.06, "cache_read": 0.03, "cache_write": 0.03 },
584
- "xopqwen35397b": { "input": 0.17, "output": 1.00, "cache_read": 0.17, "cache_write": 0.17 },
585
- "xminimaxm25": { "input": 0.29, "output": 1.17, "cache_read": 0.29, "cache_write": 0.29 },
586
- "xopglm5": { "input": 0.83, "output": 3.06, "cache_read": 0.17, "cache_write": 0.83 },
587
- "xopkimik25": { "input": 0.56, "output": 2.92, "cache_read": 0.56, "cache_write": 0.56 },
588
- "xopdeepseekv32": { "input": 0.14, "output": 0.21, "cache_read": 0.14, "cache_write": 0.14 },
589
- "xop3qwencodernext": { "input": 0.35, "output": 1.39, "cache_read": 0.35, "cache_write": 0.35 },
590
- "xopglmv47flash": { "input": 0.14, "output": 0.21, "cache_read": 0.14, "cache_write": 0.14 },
591
- "xopglm47blth2": { "input": 0.56, "output": 2.22, "cache_read": 0.56, "cache_write": 0.56 },
592
- "xop3qwen32bvl": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
593
- "xopdeepseekocr": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
594
- "xophunyuanocr": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
595
- "xop3qwen80bnext": { "input": 0.08, "output": 0.33, "cache_read": 0.08, "cache_write": 0.08 },
596
- "xop3qwen235b2507": { "input": 0.17, "output": 1.67, "cache_read": 0.17, "cache_write": 0.17 },
597
- "xop3qwen30b2507": { "input": 0.06, "output": 0.63, "cache_read": 0.06, "cache_write": 0.06 },
598
- "xop3qwen235b": { "input": 0.17, "output": 1.67, "cache_read": 0.17, "cache_write": 0.17 },
599
- "xop3qwen30b": { "input": 0.06, "output": 0.63, "cache_read": 0.06, "cache_write": 0.06 },
600
- "xop3qwen32b": { "input": 0.17, "output": 1.67, "cache_read": 0.17, "cache_write": 0.17 },
601
- "xdeepseekv3": { "input": 0.22, "output": 0.89, "cache_read": 0.22, "cache_write": 0.22 },
602
- "xdeepseekr1": { "input": 0.44, "output": 1.78, "cache_read": 0.44, "cache_write": 0.44 },
603
- "xdeepseekr1qwen32b": { "input": 0.22, "output": 0.67, "cache_read": 0.22, "cache_write": 0.22 },
604
- "xopkimik2blth": { "input": 0.56, "output": 2.22, "cache_read": 0.56, "cache_write": 0.56 },
605
- "xopkimik2blins": { "input": 0.56, "output": 2.22, "cache_read": 0.56, "cache_write": 0.56 },
606
- "xop3qwen8breranker": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
607
- "xop3qwen8bembedding": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
608
- "xop3qwen0b6": { "input": 0.04, "output": 0.42, "cache_read": 0.04, "cache_write": 0.04 },
609
- "xop3qwen4b": { "input": 0.04, "output": 0.42, "cache_read": 0.04, "cache_write": 0.04 },
610
- "xqwen257bchat": { "input": 0.07, "output": 0.14, "cache_read": 0.07, "cache_write": 0.07 },
611
- "xop3qwen14b": { "input": 0.14, "output": 1.39, "cache_read": 0.14, "cache_write": 0.14 },
612
- "xop3qwen8b": { "input": 0.07, "output": 0.69, "cache_read": 0.07, "cache_write": 0.07 },
613
- "xsparkprox": { "input": 1.11, "output": 5.56, "cache_read": 1.11, "cache_write": 1.11 },
614
- "xspark13b6k": { "input": 0.28, "output": 0.83, "cache_read": 0.28, "cache_write": 0.28 },
615
- "spark mini": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
616
- "spark mini instruct": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
617
- "spark tiny": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
618
- "internlm2.5_7b_chat": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
619
- "internlm2.5_1.8b_chat": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
620
- "qwen_v2.5_7b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
621
- "xsqwen2d53b": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
622
- "qwen_v2.5_3b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
623
- "qwen_v2.5_1.5b_instruct": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
624
- "qwen_v2.5_1.5b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
625
- "qwen_v2.5_0.5b_instruct": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
626
- "qwen_v2.5_0.5b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
627
- "xsqwenv2s1b5c": { "input": 0.14, "output": 0.28, "cache_read": 0.14, "cache_write": 0.14 },
628
- "xsqwenv2s0b5c": { "input": 0.28, "output": 0.56, "cache_read": 0.28, "cache_write": 0.28 },
629
- "xqwen14bchat": { "input": 0.28, "output": 0.83, "cache_read": 0.28, "cache_write": 0.28 }
630
- }
631
- },
632
- "fuzzy": [
633
- {
634
- "match": "cline-free/",
635
- "ref": "cline-gateway-models"
636
- },
637
- {
638
- "match": "cline-pass/",
639
- "ref": "cline-gateway-models"
640
- },
641
- {
642
- "match": "gpt-6-astra",
643
- "ref": "gpt-6-astra"
644
- },
645
- {
646
- "match": "gpt-6.1-sol",
647
- "ref": "gpt-6.1-sol"
648
- },
649
- {
650
- "match": "gpt-6-sol",
651
- "ref": "gpt-6-sol"
652
- },
653
- {
654
- "match": "gpt-5.6-sol",
655
- "ref": "gpt-5.6-sol"
656
- },
657
- {
658
- "match": "gpt-5.6-terra",
659
- "ref": "gpt-5.6-terra"
660
- },
661
- {
662
- "match": "gpt-5.6-luna",
663
- "ref": "gpt-5.6-luna"
664
- },
665
- {
666
- "match": "gpt-5.6",
667
- "ref": "gpt-5.6-sol"
668
- },
669
- {
670
- "match": "kiro",
671
- "ref": "kiro-cli-agent"
672
- },
673
- {
674
- "match": "hy3",
675
- "ref": "hy3-preview-agent"
676
- },
677
- {
678
- "match": "hy4",
679
- "ref": "hy4-preview"
680
- },
681
- {
682
- "match": "composer",
683
- "ref": "composer-1"
684
- },
685
- {
686
- "match": "claude-opus-5-fast",
687
- "ref": "claude-opus-5-fast"
688
- },
689
- {
690
- "match": "claude-opus-5",
691
- "ref": "claude-opus-5"
692
- },
693
- {
694
- "match": "claude-opus-4-8",
695
- "ref": "claude-opus-4-8"
696
- },
697
- {
698
- "match": "minimax-m3",
699
- "ref": "minimax-m3"
700
- },
701
- {
702
- "match": "minimax-m2.7-highspeed",
703
- "ref": "MiniMax-M2.7-highspeed"
704
- },
705
- {
706
- "match": "minimax-m2.7",
707
- "ref": "MiniMax-M2.7"
708
- },
709
- {
710
- "match": "deepseek-v4.1-flash",
711
- "ref": "deepseek-v4.1-flash"
712
- },
713
- {
714
- "match": "deepseek-flash",
715
- "ref": "deepseek-flash"
716
- },
717
- {
718
- "match": "deepseek-v4-flash",
719
- "ref": "deepseek-v4-flash"
720
- },
721
- {
722
- "match": "deepseek-v4-pro",
723
- "ref": "deepseek-v4-pro"
724
- },
725
- {
726
- "match": "kimi-k2.7-code",
727
- "ref": "kimi-k2.7-code"
728
- },
729
- {
730
- "match": "kimi-k3",
731
- "ref": "kimi-k3"
732
- },
733
- {
734
- "match": "kimi-k2.6",
735
- "ref": "kimi-k2.6"
736
- },
737
- {
738
- "match": "kimi",
739
- "ref": "kimi-k2.5"
740
- },
741
- {
742
- "match": "glm-4.5-airx",
743
- "ref": "glm-4.5-airx"
744
- },
745
- {
746
- "match": "glm-4.5-air",
747
- "ref": "glm-4.5-air"
748
- },
749
- {
750
- "match": "glm-4.5-x",
751
- "ref": "glm-4.5-x"
752
- },
753
- {
754
- "match": "glm-4.5-flash",
755
- "ref": "glm-4.5-flash"
756
- },
757
- {
758
- "match": "glm-4.5",
759
- "ref": "glm-4.5"
760
- },
761
- {
762
- "match": "glm-4.7-flashx",
763
- "ref": "glm-4.7-flashx"
764
- },
765
- {
766
- "match": "glm-4.7-flash",
767
- "ref": "glm-4.7-flash"
768
- },
769
- {
770
- "match": "glm-4.7",
771
- "ref": "glm-4.7"
772
- },
773
- {
774
- "match": "glm-4.6",
775
- "ref": "glm-4.6"
776
- },
777
- {
778
- "match": "glm-5.3-flash",
779
- "ref": "glm-5.3-flash"
780
- },
781
- {
782
- "match": "glm-5.3",
783
- "ref": "glm-5.3"
784
- },
785
- {
786
- "match": "glm-5-turbo",
787
- "ref": "glm-5-turbo"
788
- },
789
- {
790
- "match": "glm-5.2",
791
- "ref": "glm-5.2"
792
- },
793
- {
794
- "match": "glm-5.1",
795
- "ref": "glm-5.1"
796
- },
797
- {
798
- "match": "glm-5",
799
- "ref": "glm-5"
800
- },
801
- {
802
- "match": "fugu",
803
- "ref": "sakana/fugu-ultra"
804
- },
805
- {
806
- "match": "longcat",
807
- "ref": "longcat-2.0"
808
- },
809
- {
810
- "match": "step-3.7-flash",
811
- "ref": "step-3.7-flash"
812
- },
813
- {
814
- "match": "step-3.5-flash",
815
- "ref": "step-3.5-flash"
816
- },
817
- {
818
- "match": "stepfun",
819
- "ref": "step-3.7-flash"
820
- }
821
- ]
822
- }
1
+ {
2
+ "_meta": {
3
+ "note": "Curated price overrides. Always wins over LiteLLM. Two reasons to live here: (1) self-defined alias names that LiteLLM will never carry (kiro-*, hy3-*, composer-*, kimi-for-coding, free-tier OpenRouter routes); (2) prices we want to pin even if LiteLLM has the model (e.g. cache_write fields LiteLLM often omits). Units: USD per million tokens. Edit this file to override pricing without redeploying.",
4
+ "units": "usd_per_million_tokens",
5
+ "deepseek_v4_pricing_note": "Official DeepSeek V4 pricing verified on 2026-08-21. Peak hours are 01:00-04:00 and 06:00-10:00 UTC; all other hours are off-peak at half price. Exact entries store peak rates. computeRowCost applies the time-of-use discount from each row's UTC hour_start.",
6
+ "iflytek_maas_pricing_note": "iFlytek MaaS inference prices supplied for AStudio on 2026-08-31. AStudio homepage: https://agent.xfyun.cn/. Official pricing source: https://maas.xfyun.cn/modelSquare. Source values are RMB per million tokens, converted at 7.2 RMB/USD, then rounded to two decimal places for billing. Models without a dedicated cache-hit price use the normal input rate; cache writes also use the normal input rate. iFlytek MaaS prices are source-scoped and do not inherit DeepSeek time-of-use discounts. Spark-X2.5 entries (spark-x2.5, spark-x2.5-4b, spark-x2.5-1.7b) were read from the model square on 2026-10-08; spark-x2.5 is listed at a limited-time 30% rate (0.96 / 3.6 / 0.14 RMB input / output / cache-hit, list price 3.2 / 12 / 0.48), and the two small models are free."
7
+ },
8
+ "exact": {
9
+ "claude-opus-5-5": {
10
+ "input": 4,
11
+ "output": 20,
12
+ "cache_read": 0.2,
13
+ "cache_write": 5,
14
+ "note": "Official Anthropic pricing verified 2026-10-11: https://platform.claude.com/docs/en/about-claude/pricing"
15
+ },
16
+ "claude-opus-5.5": {
17
+ "input": 4,
18
+ "output": 20,
19
+ "cache_read": 0.2,
20
+ "cache_write": 5,
21
+ "note": "Official Anthropic pricing verified 2026-10-11: https://platform.claude.com/docs/en/about-claude/pricing"
22
+ },
23
+ "claude-sonnet-5-5": {
24
+ "input": 2,
25
+ "output": 10,
26
+ "cache_read": 0.1,
27
+ "cache_write": 2.5,
28
+ "note": "Official Anthropic pricing verified 2026-10-11: https://platform.claude.com/docs/en/about-claude/pricing"
29
+ },
30
+ "claude-sonnet-5.5": {
31
+ "input": 2,
32
+ "output": 10,
33
+ "cache_read": 0.1,
34
+ "cache_write": 2.5,
35
+ "note": "Official Anthropic pricing verified 2026-10-11: https://platform.claude.com/docs/en/about-claude/pricing"
36
+ },
37
+ "k3-256k": {
38
+ "input": 3,
39
+ "output": 15,
40
+ "cache_read": 0.3,
41
+ "note": "Bare alias emitted by Kimi Code (modelAlias \"kimi-code/k3\" -> \"k3\"). Same rates as kimi-k3."
42
+ },
43
+ "cline-gateway-models": {
44
+ "input": 0,
45
+ "output": 0,
46
+ "cache_read": 0,
47
+ "cache_write": 0,
48
+ "note": "Cline's own gateway namespaces (`cline-free/*` free tier, `cline-pass/*` flat-rate Cline Pass). Cline's shipped model catalog declares pricing 0 for every id in both namespaces, so a turn there has no marginal per-token cost — same convention as PI_SUBSCRIPTION_SOURCES. The fuzzy entries at the top of `fuzzy` route both prefixes here BEFORE the model names can match a public rate (`cline-pass/glm-5.3` must not bill GLM-5.3 list price, `cline-free/deepseek-v4.1-flash` must not bill DeepSeek). Verified against Cline 0.0.32 (/Applications/Cline.app model catalog + `usageDelta` cost, which stays 0 for both namespaces). A Cline turn that DOES report a positive per-call cost is honoured first because `cline` is listed in SOURCES_WITH_AUTHORITATIVE_COST in pricing/index.js. Remove or re-point these entries when Cline publishes per-token pricing for either namespace."
49
+ },
50
+ "claude-fable-5": {
51
+ "input": 10,
52
+ "output": 50,
53
+ "cache_read": 1,
54
+ "cache_write": 12.5,
55
+ "note": "Fable 5 \u2014 Anthropic's top tier above Opus. Not yet in LiteLLM. $10/$50 per MTok; cache_read 0.1x, cache_write 1.25x. Remove once LiteLLM carries it."
56
+ },
57
+ "claude-opus-5": {
58
+ "input": 5,
59
+ "output": 25,
60
+ "cache_read": 0.5,
61
+ "cache_write": 6.25,
62
+ "note": "Opus 5 (public 2026-07). Standard Opus tier, unchanged from 4.6/4.7/4.8: $5/$25 per MTok, cache_read 0.1x, cache_write 1.25x. Not yet in LiteLLM. Fast mode is billed at $10/$50 \u2014 see claude-opus-5-fast. Remove once LiteLLM carries it."
63
+ },
64
+ "claude-opus-5-fast": {
65
+ "input": 10,
66
+ "output": 50,
67
+ "cache_read": 1,
68
+ "cache_write": 12.5,
69
+ "note": "Opus 5 fast mode (speed=\"fast\", beta fast-mode-2026-02-01) is priced at $10/$50 per MTok \u2014 2x standard Opus. Kept separate so a -fast model id is not silently priced at the standard tier. See claude-opus-5."
70
+ },
71
+ "claude-opus-4-8": {
72
+ "input": 5,
73
+ "output": 25,
74
+ "cache_read": 0.5,
75
+ "cache_write": 6.25,
76
+ "note": "Opus 4.8 not yet in LiteLLM. Pin to the standard Opus tier (same as 4.6/4.7); remove once LiteLLM carries it."
77
+ },
78
+ "claude-sonnet-5": {
79
+ "input": 2,
80
+ "output": 10,
81
+ "cache_read": 0.2,
82
+ "cache_write": 2.5,
83
+ "note": "Sonnet 5. LiteLLM added this model on ~2026-06-30 with matching rates, but pin explicitly so pricing is correct immediately rather than depending on the 24h local pricing-cache refresh cycle."
84
+ },
85
+ "claude-haiku-4-5": {
86
+ "input": 1,
87
+ "output": 5,
88
+ "cache_read": 0.1,
89
+ "cache_write": 1.25,
90
+ "note": "Canonical undated alias emitted by Cursor. Pin to the same rates as claude-haiku-4-5-20251001 so local cost does not depend on which dated LiteLLM aliases are present in the current cache."
91
+ },
92
+ "gpt-6-astra": {
93
+ "input": 10,
94
+ "output": 50,
95
+ "cache_read": 1,
96
+ "cache_write": 12.5,
97
+ "priority_multiplier": 2,
98
+ "note": "Official Standard USD/MTok rates verified 2026-09-07: https://developers.openai.com/api/docs/models/gpt-6-astra. Requests above 272K input tokens use 2x input/cache and 1.5x output rates; computeRowCost applies this only to observed long-context request subsets. priority_multiplier is the Fast tier, whose published table (20/2/25/100 short, 40/4/50/150 long, verified 2026-09-12) is exactly 2x the Standard rate of the same context tier — it multiplies, it does not replace, the long-context premium. The 2.5x figure in the Codex speed docs is a subscription-credit rate and must not be used here. computeRowCost applies it only to observed priority request subsets; usage whose service tier the logs never recorded stays Standard. Cloud aggregates use Standard short-context estimates because they do not retain per-request context or service tier. Reasoning-effort suffixes use the same base rates."
99
+ },
100
+ "gpt-6.1-sol": {
101
+ "input": 2,
102
+ "output": 10,
103
+ "cache_read": 0.1,
104
+ "cache_write": 2.5,
105
+ "priority_multiplier": 2,
106
+ "note": "Official Standard USD/MTok rates verified 2026-10-02 (issue #737): https://developers.openai.com/api/docs/models/gpt-6.1-sol. Requests above 272K input use 2x input/cache and 1.5x output rates; Fast mode is 2x the applicable rates. Premiums apply only to observed request subsets. Cloud aggregates remain Standard short-context estimates."
107
+ },
108
+ "gpt-6-sol": {
109
+ "input": 2,
110
+ "output": 10,
111
+ "cache_read": 0.2,
112
+ "cache_write": 2.5,
113
+ "priority_multiplier": 2,
114
+ "note": "Official Standard USD/MTok rates verified 2026-09-24: https://developers.openai.com/api/docs/models/gpt-6-sol. Same rules as gpt-6-astra: requests above 272K input tokens use 2x input/cache and 1.5x output for the full request, applied only to observed long-context request subsets; Fast mode is published as 2x the applicable rates (priority_multiplier). Reasoning-effort suffixes use the same base rates."
115
+ },
116
+ "gpt-5.6-sol": {
117
+ "input": 4,
118
+ "output": 20,
119
+ "cache_read": 0.4,
120
+ "cache_write": 5,
121
+ "note": "Official GPT-5.6 Sol standard pricing: $4 input, $0.40 cached input, $5 cache write, and $20 output per MTok. Requests above 272K input are repriced from their preserved per-request long-context subset. Codex emits gpt-5.6-sol plus reasoning-effort variants caught by fuzzy matching."
122
+ },
123
+ "gpt-5.6-terra": {
124
+ "input": 2,
125
+ "output": 12,
126
+ "cache_read": 0.2,
127
+ "cache_write": 2.5,
128
+ "note": "GPT-5.6 balanced default tier. Official short-context rates: $2/$12 per MTok (not the earlier $2.5/$15 snapshot). See gpt-5.6-sol."
129
+ },
130
+ "gpt-5.6-luna": {
131
+ "input": 0.2,
132
+ "output": 1.2,
133
+ "cache_read": 0.02,
134
+ "cache_write": 0.25,
135
+ "note": "GPT-5.6 lightweight tier. Official short-context rates: $0.20/$1.20 per MTok (previous curated values $1/$6 overcounted 5x). See gpt-5.6-sol."
136
+ },
137
+ "kiro-agent": {
138
+ "input": 3,
139
+ "output": 15,
140
+ "cache_read": 0.3,
141
+ "cache_write": 3.75
142
+ },
143
+ "kiro-cli-agent": {
144
+ "input": 3,
145
+ "output": 15,
146
+ "cache_read": 0.3,
147
+ "cache_write": 3.75
148
+ },
149
+ "hy3-preview-agent": {
150
+ "input": 0.167,
151
+ "output": 0.556,
152
+ "cache_read": 0.056,
153
+ "cache_write": 0.167,
154
+ "note": "Tencent Hunyuan Hy3 preview (CodeBuddy/WorkBuddy backend). Official TokenHub rate: 1.2 / 0.4 (cache hit) / 4.0 RMB per MTok in/read/out, converted at ~7.2 RMB/USD. DeepSeek-style cache: no write surcharge, so cache_write = input."
155
+ },
156
+ "hy3-preview": {
157
+ "input": 0.167,
158
+ "output": 0.556,
159
+ "cache_read": 0.056,
160
+ "cache_write": 0.167,
161
+ "note": "See hy3-preview-agent."
162
+ },
163
+ "qwen3.8-flash": {
164
+ "input": 0.15,
165
+ "output": 0.47,
166
+ "cache_read": 0.016,
167
+ "cache_write": 0.2,
168
+ "note": "Alibaba Model Studio Singapore (International), USD per MTok, verified 2026-09-29: https://www.alibabacloud.com/help/en/model-studio/qwen3-8-flash . Reference API estimate; other regions/providers can have different rates. Issue #715."
169
+ },
170
+ "hy4-preview": {
171
+ "input": 0.833,
172
+ "output": 2.5,
173
+ "cache_read": 0.042,
174
+ "cache_write": 0.833,
175
+ "note": "Tencent Hunyuan Hy4 preview (released 2026-08-28; WorkBuddy/CodeBuddy backend). Official TokenHub rate: 6 / 0.3 (cache hit) / 18 RMB per MTok in/read/out, converted at ~7.2 RMB/USD like hy3-preview. No published cache-write price, so cache_write = input. Issue #633."
176
+ },
177
+ "hy4-preview-agent": {
178
+ "input": 0.833,
179
+ "output": 2.5,
180
+ "cache_read": 0.042,
181
+ "cache_write": 0.833,
182
+ "note": "See hy4-preview. CodeBuddy's auto-router emits the -agent suffixed form."
183
+ },
184
+ "composer-1": {
185
+ "input": 1.25,
186
+ "output": 10,
187
+ "cache_read": 0.125
188
+ },
189
+ "composer-1.5": {
190
+ "input": 3.5,
191
+ "output": 17.5,
192
+ "cache_read": 0.35
193
+ },
194
+ "composer-2": {
195
+ "input": 0.5,
196
+ "output": 2.5,
197
+ "cache_read": 0.2
198
+ },
199
+ "composer-2-fast": {
200
+ "input": 1.5,
201
+ "output": 7.5,
202
+ "cache_read": 0.15
203
+ },
204
+ "composer-2.5": {
205
+ "input": 0.5,
206
+ "output": 2.5,
207
+ "cache_read": 0.2,
208
+ "note": "Cursor Composer 2.5. Cursor Models & Pricing: $0.50/M input, $0.20/M cached input, $2.50/M output; no cache-write rate is published. https://cursor.com/docs/models-and-pricing"
209
+ },
210
+ "composer-2.5-fast": {
211
+ "input": 3.0,
212
+ "output": 15.0,
213
+ "cache_read": 0.5,
214
+ "note": "Cursor Composer 2.5 Fast. Cursor Models & Pricing: $3/M input, $0.50/M cached input, $15/M output. Kept separate from the cheaper base SKU."
215
+ },
216
+ "gemini-4-argon": {
217
+ "input": 2.0,
218
+ "output": 10.0,
219
+ "cache_read": 0.1,
220
+ "cache_write": 0,
221
+ "note": "Google Gemini 4 Argon introductory list rate (through 2026-11-30) matching Cursor Other Models pool: $2/M input, $0.10/M cached input, $10/M output. Effort suffixes (-high/-medium/-low/-minimal) are stripped by normalizeCursorModel. Sources: https://cursor.com/docs/models-and-pricing and https://ai.google.dev/gemini-api/docs/pricing"
222
+ },
223
+ "MiniMax-M2.7": {
224
+ "input": 0.3,
225
+ "output": 1.2,
226
+ "cache_read": 0.06,
227
+ "cache_write": 0.375
228
+ },
229
+ "MiniMax-M2.7-highspeed": {
230
+ "input": 0.6,
231
+ "output": 2.4,
232
+ "cache_read": 0.06,
233
+ "cache_write": 0.375
234
+ },
235
+ "minimax-m3": {
236
+ "input": 0.3,
237
+ "output": 1.2,
238
+ "cache_read": 0.06,
239
+ "cache_write": 0
240
+ },
241
+ "deepseek-v4-flash": {
242
+ "input": 0.44,
243
+ "output": 1.32,
244
+ "cache_read": 0.014,
245
+ "cache_write": 0.44
246
+ },
247
+ "deepseek-v4-pro": {
248
+ "input": 1.32,
249
+ "output": 3.96,
250
+ "cache_read": 0.044,
251
+ "cache_write": 1.32
252
+ },
253
+ "deepseek-v4-flash-vision-exp": {
254
+ "input": 0.44,
255
+ "output": 1.32,
256
+ "cache_read": 0.014,
257
+ "cache_write": 0.44
258
+ },
259
+ "deepseek-v4.1-flash": {
260
+ "input": 0.3,
261
+ "output": 1.2,
262
+ "cache_read": 0.006,
263
+ "cache_write": 0.3,
264
+ "note": "DeepSeek V4.1 Flash (official API id deepseek-flash, released 2026-09-10). deepseek-v4.1-flash is the id reported by OpenRouter, Command Code and WorkBuddy. Peak rates per https://api-docs.deepseek.com/quick_start/pricing/ ($0.30 / $1.20 / $0.006 cache read per MTok); computeRowCost halves them off-peak like the other DeepSeek V4 entries."
265
+ },
266
+ "deepseek-flash": {
267
+ "input": 0.3,
268
+ "output": 1.2,
269
+ "cache_read": 0.006,
270
+ "cache_write": 0.3,
271
+ "note": "Official DeepSeek API id for DeepSeek-V4.1-Flash. Same rates as deepseek-v4.1-flash."
272
+ },
273
+ "deepseek-chat": {
274
+ "input": 0.14,
275
+ "output": 0.28,
276
+ "cache_read": 0.0028,
277
+ "cache_write": 0.14
278
+ },
279
+ "grok-build": {
280
+ "input": 1.25,
281
+ "output": 2.5,
282
+ "cache_read": 0.2,
283
+ "note": "Grok Build TUI fallback estimate for sessions without turn_completed.usage. Preferred path is turn_completed usage with per-model modelUsage splits."
284
+ },
285
+ "cursor-grok-4.5": {
286
+ "input": 2.0,
287
+ "output": 6.0,
288
+ "cache_read": 0.5,
289
+ "cache_write": 0,
290
+ "note": "Cursor Grok 4.5 base rate. Cursor Models & Pricing: $2/M input, $0.50/M cached input, $6/M output; no cache-write rate is published."
291
+ },
292
+ "cursor-grok-4.5-fast": {
293
+ "input": 4.0,
294
+ "output": 18.0,
295
+ "cache_read": 1.0,
296
+ "cache_write": 0,
297
+ "note": "Cursor Grok 4.5 Fast rate. Cursor publishes $4/M input and $18/M output; the Cursor usage breakdown reports $1/M cached input. Kept separate from the cheaper base SKU."
298
+ },
299
+ "cursor-grok-4.6": {
300
+ "input": 2.0,
301
+ "output": 6.0,
302
+ "cache_read": 0.5,
303
+ "cache_write": 0,
304
+ "note": "Cursor Grok 4.6 base rate. Cursor Models & Pricing: $2/M input, $0.50/M cached input, $6/M output; no cache-write rate is published."
305
+ },
306
+ "cursor-grok-4.6-fast": {
307
+ "input": 4.0,
308
+ "output": 12.0,
309
+ "cache_read": 1.0,
310
+ "cache_write": 0,
311
+ "note": "Cursor Grok 4.6 Fast rate. Cursor publishes $4/M input, $1/M cached input, $12/M output. Kept separate from the cheaper base SKU."
312
+ },
313
+ "cursor-grok-4.7": {
314
+ "input": 2.0,
315
+ "output": 6.0,
316
+ "cache_read": 0.5,
317
+ "cache_write": 0,
318
+ "note": "Cursor Grok 4.7 base rate. Cursor Models & Pricing: $2/M input, $0.50/M cached input, $6/M output. Long-context (>256k) premiums are not applied here yet because Cursor CSV rows do not expose per-request long-context flags."
319
+ },
320
+ "cursor-grok-4.7-fast": {
321
+ "input": 4.0,
322
+ "output": 12.0,
323
+ "cache_read": 1.0,
324
+ "cache_write": 0,
325
+ "note": "Cursor Grok 4.7 Fast rate. Cursor publishes $4/M input, $1/M cached input, $12/M output. Kept separate from the cheaper base SKU."
326
+ },
327
+ "grok-4-0709": {
328
+ "input": 3.0,
329
+ "output": 15.0,
330
+ "cache_read": 0.75
331
+ },
332
+ "grok-4": {
333
+ "input": 3.0,
334
+ "output": 15.0,
335
+ "cache_read": 0.75
336
+ },
337
+ "grok-4-latest": {
338
+ "input": 3.0,
339
+ "output": 15.0,
340
+ "cache_read": 0.75
341
+ },
342
+ "grok-4-fast": {
343
+ "input": 0.2,
344
+ "output": 0.5,
345
+ "cache_read": 0.05
346
+ },
347
+ "grok-4-fast-reasoning": {
348
+ "input": 0.2,
349
+ "output": 0.5,
350
+ "cache_read": 0.05
351
+ },
352
+ "grok-4-fast-non-reasoning": {
353
+ "input": 0.2,
354
+ "output": 0.5,
355
+ "cache_read": 0.05
356
+ },
357
+ "grok-4-1-fast-non-reasoning": {
358
+ "input": 0.2,
359
+ "output": 0.5,
360
+ "cache_read": 0.05
361
+ },
362
+ "deepseek-reasoner": {
363
+ "input": 0.14,
364
+ "output": 0.28,
365
+ "cache_read": 0.0028,
366
+ "cache_write": 0.14
367
+ },
368
+ "kimi-for-coding": {
369
+ "input": 0.6,
370
+ "output": 2,
371
+ "cache_read": 0.15
372
+ },
373
+ "kimi-k2.5": {
374
+ "input": 0.6,
375
+ "output": 2,
376
+ "cache_read": 0.15
377
+ },
378
+ "kimi-k2.5-free": {
379
+ "input": 0,
380
+ "output": 0,
381
+ "cache_read": 0
382
+ },
383
+ "kimi-k2.6": {
384
+ "input": 0.95,
385
+ "output": 4,
386
+ "cache_read": 0.16
387
+ },
388
+ "kimi-k2.7-code": {
389
+ "input": 0.95,
390
+ "output": 4,
391
+ "cache_read": 0.19
392
+ },
393
+ "kimi-k3": {
394
+ "input": 3,
395
+ "output": 15,
396
+ "cache_read": 0.3,
397
+ "note": "Kimi K3 (released 2026-07-16). Reported API rates: $3/M input, $15/M output, $0.30/M cached input (0.1x). Not yet in LiteLLM; remove once it carries k3. Kimi Code records the alias as bare \"k3\" (kimi-code/k3), hence the separate exact key below."
398
+ },
399
+ "k3": {
400
+ "input": 3,
401
+ "output": 15,
402
+ "cache_read": 0.3,
403
+ "note": "Bare alias emitted by Kimi Code (modelAlias \"kimi-code/k3\" -> \"k3\"). Same rates as kimi-k3."
404
+ },
405
+ "glm-5.3": {
406
+ "input": 1.4,
407
+ "output": 4.4,
408
+ "cache_read": 0.26
409
+ },
410
+ "glm-5.3-flash": {
411
+ "input": 0.15,
412
+ "output": 0.5,
413
+ "cache_read": 0.03
414
+ },
415
+ "glm-5.2": {
416
+ "input": 1.4,
417
+ "output": 4.4,
418
+ "cache_read": 0.26
419
+ },
420
+ "glm-5.1": {
421
+ "input": 1.4,
422
+ "output": 4.4,
423
+ "cache_read": 0.26
424
+ },
425
+ "glm-5": {
426
+ "input": 1.0,
427
+ "output": 3.2,
428
+ "cache_read": 0.2
429
+ },
430
+ "glm-5-turbo": {
431
+ "input": 1.2,
432
+ "output": 4.0,
433
+ "cache_read": 0.24
434
+ },
435
+ "glm-4.7": {
436
+ "input": 0.6,
437
+ "output": 2.2,
438
+ "cache_read": 0.11
439
+ },
440
+ "glm-4.7-flashx": {
441
+ "input": 0.07,
442
+ "output": 0.4,
443
+ "cache_read": 0.01
444
+ },
445
+ "glm-4.7-flash": {
446
+ "input": 0,
447
+ "output": 0,
448
+ "cache_read": 0
449
+ },
450
+ "glm-4.6": {
451
+ "input": 0.6,
452
+ "output": 2.2,
453
+ "cache_read": 0.11
454
+ },
455
+ "glm-4.5": {
456
+ "input": 0.6,
457
+ "output": 2.2,
458
+ "cache_read": 0.11
459
+ },
460
+ "glm-4.5-x": {
461
+ "input": 2.2,
462
+ "output": 8.9,
463
+ "cache_read": 0.45
464
+ },
465
+ "glm-4.5-air": {
466
+ "input": 0.2,
467
+ "output": 1.1,
468
+ "cache_read": 0.03
469
+ },
470
+ "glm-4.5-airx": {
471
+ "input": 1.1,
472
+ "output": 4.5,
473
+ "cache_read": 0.22
474
+ },
475
+ "glm-4.5-flash": {
476
+ "input": 0,
477
+ "output": 0,
478
+ "cache_read": 0
479
+ },
480
+ "glm-4.7-free": {
481
+ "input": 0,
482
+ "output": 0,
483
+ "cache_read": 0
484
+ },
485
+ "nemotron-3-super-free": {
486
+ "input": 0,
487
+ "output": 0,
488
+ "cache_read": 0
489
+ },
490
+ "mimo-v2-pro-free": {
491
+ "input": 0,
492
+ "output": 0,
493
+ "cache_read": 0
494
+ },
495
+ "minimax-m2.1-free": {
496
+ "input": 0,
497
+ "output": 0,
498
+ "cache_read": 0
499
+ },
500
+ "MiniMax-M2.1": {
501
+ "input": 0.5,
502
+ "output": 3,
503
+ "cache_read": 0.05
504
+ },
505
+ "antigravity-gpt-oss-120b": {
506
+ "input": 2.5,
507
+ "output": 10,
508
+ "cache_read": 0,
509
+ "note": "Antigravity bridge alias. Approximate with GPT-4o-class pricing until Google exposes route-specific billing metadata."
510
+ },
511
+ "sakana/fugu-ultra": {
512
+ "input": 5,
513
+ "output": 30,
514
+ "cache_read": 0.5,
515
+ "cache_write": 5,
516
+ "note": "Sakana Fugu Ultra \u2014 multi-agent orchestration via an OpenAI-compatible API (sakana.ai PAYG + OpenRouter), used through Codex/Cursor/Cline/ZCode etc. OpenRouter rates: $5/$30 per MTok in/out, cache_read $0.5/M; no published cache-write surcharge, so cache_write = input. Subscriptions ($20/$100/$200) bill against this token-equivalent rate."
517
+ },
518
+ "longcat-2.0": {
519
+ "input": 0.278,
520
+ "output": 1.111,
521
+ "cache_read": 0.00556,
522
+ "cache_write": 0.278,
523
+ "note": "Meituan LongCat-2.0, seen via ZCode custom-provider routing (issue #276). Official longcat.chat/platform launch-promo rate: RMB 2 / 0.04 (cache hit) / 8 per MTok in/read/out, converted at ~7.2 RMB/USD. Standard list price is RMB 5/0.10/20 (2.5x) once the promo ends \u2014 re-verify against longcat.chat/platform/docs/zh/Pricing/LongCat-2.0.html before assuming either is permanent. DeepSeek-style cache: no write surcharge, so cache_write = input."
524
+ },
525
+ "step-3.7-flash": {
526
+ "input": 0.2,
527
+ "output": 1.15,
528
+ "cache_read": 0.04,
529
+ "cache_write": 0.2,
530
+ "note": "StepFun Step 3.7 Flash (issue #283). Official platform.stepfun.ai pricing: $0.20 / $0.04 (cache hit) / $1.15 per MTok in/read/out. No published cache-write surcharge, so cache_write = input."
531
+ },
532
+ "step-3.5-flash": {
533
+ "input": 0.1,
534
+ "output": 0.3,
535
+ "cache_read": 0.02,
536
+ "cache_write": 0.1,
537
+ "note": "StepFun Step 3.5 Flash incl. dated snapshots like step-3.5-flash-2603 (issue #283). Official platform.stepfun.ai pricing: $0.10 / $0.02 (cache hit) / $0.30 per MTok in/read/out. No published cache-write surcharge, so cache_write = input."
538
+ },
539
+ "grok-4.5-build": {
540
+ "input": 2.0,
541
+ "output": 6.0,
542
+ "cache_read": 0.5,
543
+ "cache_write": 0,
544
+ "note": "Grok Build paid SKU (grok-4.5-build from turn_completed.modelUsage). Rates aligned with xAI/Cursor Grok 4.5 until a dedicated Build price list is published."
545
+ },
546
+ "grok-build-free": {
547
+ "input": 0,
548
+ "output": 0,
549
+ "cache_read": 0,
550
+ "cache_write": 0,
551
+ "note": "Grok Build free tier (canonical model id). $0 marginal cost. Prefer this over grok-4.5-build-free so pricing lookup cannot fuzzy-match paid grok-4.5."
552
+ },
553
+ "grok-4.5-build-free": {
554
+ "input": 0,
555
+ "output": 0,
556
+ "cache_read": 0,
557
+ "cache_write": 0,
558
+ "note": "Alias for free Build SKU as labeled in Grok updates.jsonl modelUsage. Canonical storage id is grok-build-free."
559
+ }
560
+ },
561
+ "alias": {
562
+ "auto": "composer-1"
563
+ },
564
+ "source_exact": {
565
+ "acode": {
566
+ "spark-x2.5": { "input": 0.13, "output": 0.50, "cache_read": 0.02, "cache_write": 0.13 },
567
+ "spark-x2.5-4b": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
568
+ "spark-x2.5-1.7b": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
569
+ "xopglm53": { "input": 1.11, "output": 3.89, "cache_read": 0.28, "cache_write": 1.11 },
570
+ "xopdeepseekv4pro0813": { "input": 1.25, "output": 3.75, "cache_read": 0.04, "cache_write": 1.25 },
571
+ "xopdeepseekv4flash0731": { "input": 0.14, "output": 0.28, "cache_read": 0.03, "cache_write": 0.14 },
572
+ "xopkimik27code": { "input": 0.90, "output": 3.75, "cache_read": 0.90, "cache_write": 0.90 },
573
+ "xopglm52": { "input": 1.11, "output": 3.89, "cache_read": 0.28, "cache_write": 1.11 },
574
+ "xopdeepseekv4flash": { "input": 0.14, "output": 0.28, "cache_read": 0.03, "cache_write": 0.14 },
575
+ "xopkimik26": { "input": 0.90, "output": 3.75, "cache_read": 0.18, "cache_write": 0.90 },
576
+ "xopdeepseekv4pro": { "input": 1.67, "output": 3.33, "cache_read": 0.14, "cache_write": 1.67 },
577
+ "xopqwen36v35b": { "input": 0.15, "output": 0.90, "cache_read": 0.15, "cache_write": 0.15 },
578
+ "xophunyuan7bmt": { "input": 0.07, "output": 0.28, "cache_read": 0.07, "cache_write": 0.07 },
579
+ "xoppaddleocrv16": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
580
+ "xsparkx2flash": { "input": 0.14, "output": 0.28, "cache_read": 0.14, "cache_write": 0.14 },
581
+ "xopglm51": { "input": 1.11, "output": 3.89, "cache_read": 0.22, "cache_write": 1.11 },
582
+ "xsparkx2": { "input": 0.42, "output": 0.42, "cache_read": 0.42, "cache_write": 0.42 },
583
+ "xop35qwen2b": { "input": 0.03, "output": 0.06, "cache_read": 0.03, "cache_write": 0.03 },
584
+ "xopqwen35397b": { "input": 0.17, "output": 1.00, "cache_read": 0.17, "cache_write": 0.17 },
585
+ "xminimaxm25": { "input": 0.29, "output": 1.17, "cache_read": 0.29, "cache_write": 0.29 },
586
+ "xopglm5": { "input": 0.83, "output": 3.06, "cache_read": 0.17, "cache_write": 0.83 },
587
+ "xopkimik25": { "input": 0.56, "output": 2.92, "cache_read": 0.56, "cache_write": 0.56 },
588
+ "xopdeepseekv32": { "input": 0.14, "output": 0.21, "cache_read": 0.14, "cache_write": 0.14 },
589
+ "xop3qwencodernext": { "input": 0.35, "output": 1.39, "cache_read": 0.35, "cache_write": 0.35 },
590
+ "xopglmv47flash": { "input": 0.14, "output": 0.21, "cache_read": 0.14, "cache_write": 0.14 },
591
+ "xopglm47blth2": { "input": 0.56, "output": 2.22, "cache_read": 0.56, "cache_write": 0.56 },
592
+ "xop3qwen32bvl": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
593
+ "xopdeepseekocr": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
594
+ "xophunyuanocr": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
595
+ "xop3qwen80bnext": { "input": 0.08, "output": 0.33, "cache_read": 0.08, "cache_write": 0.08 },
596
+ "xop3qwen235b2507": { "input": 0.17, "output": 1.67, "cache_read": 0.17, "cache_write": 0.17 },
597
+ "xop3qwen30b2507": { "input": 0.06, "output": 0.63, "cache_read": 0.06, "cache_write": 0.06 },
598
+ "xop3qwen235b": { "input": 0.17, "output": 1.67, "cache_read": 0.17, "cache_write": 0.17 },
599
+ "xop3qwen30b": { "input": 0.06, "output": 0.63, "cache_read": 0.06, "cache_write": 0.06 },
600
+ "xop3qwen32b": { "input": 0.17, "output": 1.67, "cache_read": 0.17, "cache_write": 0.17 },
601
+ "xdeepseekv3": { "input": 0.22, "output": 0.89, "cache_read": 0.22, "cache_write": 0.22 },
602
+ "xdeepseekr1": { "input": 0.44, "output": 1.78, "cache_read": 0.44, "cache_write": 0.44 },
603
+ "xdeepseekr1qwen32b": { "input": 0.22, "output": 0.67, "cache_read": 0.22, "cache_write": 0.22 },
604
+ "xopkimik2blth": { "input": 0.56, "output": 2.22, "cache_read": 0.56, "cache_write": 0.56 },
605
+ "xopkimik2blins": { "input": 0.56, "output": 2.22, "cache_read": 0.56, "cache_write": 0.56 },
606
+ "xop3qwen8breranker": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
607
+ "xop3qwen8bembedding": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
608
+ "xop3qwen0b6": { "input": 0.04, "output": 0.42, "cache_read": 0.04, "cache_write": 0.04 },
609
+ "xop3qwen4b": { "input": 0.04, "output": 0.42, "cache_read": 0.04, "cache_write": 0.04 },
610
+ "xqwen257bchat": { "input": 0.07, "output": 0.14, "cache_read": 0.07, "cache_write": 0.07 },
611
+ "xop3qwen14b": { "input": 0.14, "output": 1.39, "cache_read": 0.14, "cache_write": 0.14 },
612
+ "xop3qwen8b": { "input": 0.07, "output": 0.69, "cache_read": 0.07, "cache_write": 0.07 },
613
+ "xsparkprox": { "input": 1.11, "output": 5.56, "cache_read": 1.11, "cache_write": 1.11 },
614
+ "xspark13b6k": { "input": 0.28, "output": 0.83, "cache_read": 0.28, "cache_write": 0.28 },
615
+ "spark mini": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
616
+ "spark mini instruct": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
617
+ "spark tiny": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
618
+ "internlm2.5_7b_chat": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
619
+ "internlm2.5_1.8b_chat": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
620
+ "qwen_v2.5_7b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
621
+ "xsqwen2d53b": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
622
+ "qwen_v2.5_3b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
623
+ "qwen_v2.5_1.5b_instruct": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
624
+ "qwen_v2.5_1.5b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
625
+ "qwen_v2.5_0.5b_instruct": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
626
+ "qwen_v2.5_0.5b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
627
+ "xsqwenv2s1b5c": { "input": 0.14, "output": 0.28, "cache_read": 0.14, "cache_write": 0.14 },
628
+ "xsqwenv2s0b5c": { "input": 0.28, "output": 0.56, "cache_read": 0.28, "cache_write": 0.28 },
629
+ "xqwen14bchat": { "input": 0.28, "output": 0.83, "cache_read": 0.28, "cache_write": 0.28 }
630
+ }
631
+ },
632
+ "fuzzy": [
633
+ {
634
+ "match": "cline-free/",
635
+ "ref": "cline-gateway-models"
636
+ },
637
+ {
638
+ "match": "cline-pass/",
639
+ "ref": "cline-gateway-models"
640
+ },
641
+ {
642
+ "match": "gpt-6-astra",
643
+ "ref": "gpt-6-astra"
644
+ },
645
+ {
646
+ "match": "gpt-6.1-sol",
647
+ "ref": "gpt-6.1-sol"
648
+ },
649
+ {
650
+ "match": "gpt-6-sol",
651
+ "ref": "gpt-6-sol"
652
+ },
653
+ {
654
+ "match": "gpt-5.6-sol",
655
+ "ref": "gpt-5.6-sol"
656
+ },
657
+ {
658
+ "match": "gpt-5.6-terra",
659
+ "ref": "gpt-5.6-terra"
660
+ },
661
+ {
662
+ "match": "gpt-5.6-luna",
663
+ "ref": "gpt-5.6-luna"
664
+ },
665
+ {
666
+ "match": "gpt-5.6",
667
+ "ref": "gpt-5.6-sol"
668
+ },
669
+ {
670
+ "match": "kiro",
671
+ "ref": "kiro-cli-agent"
672
+ },
673
+ {
674
+ "match": "hy3",
675
+ "ref": "hy3-preview-agent"
676
+ },
677
+ {
678
+ "match": "hy4",
679
+ "ref": "hy4-preview"
680
+ },
681
+ {
682
+ "match": "composer",
683
+ "ref": "composer-1"
684
+ },
685
+ {
686
+ "match": "claude-opus-5-fast",
687
+ "ref": "claude-opus-5-fast"
688
+ },
689
+ {
690
+ "match": "claude-opus-5",
691
+ "ref": "claude-opus-5"
692
+ },
693
+ {
694
+ "match": "claude-opus-4-8",
695
+ "ref": "claude-opus-4-8"
696
+ },
697
+ {
698
+ "match": "minimax-m3",
699
+ "ref": "minimax-m3"
700
+ },
701
+ {
702
+ "match": "minimax-m2.7-highspeed",
703
+ "ref": "MiniMax-M2.7-highspeed"
704
+ },
705
+ {
706
+ "match": "minimax-m2.7",
707
+ "ref": "MiniMax-M2.7"
708
+ },
709
+ {
710
+ "match": "deepseek-v4.1-flash",
711
+ "ref": "deepseek-v4.1-flash"
712
+ },
713
+ {
714
+ "match": "deepseek-flash",
715
+ "ref": "deepseek-flash"
716
+ },
717
+ {
718
+ "match": "deepseek-v4-flash",
719
+ "ref": "deepseek-v4-flash"
720
+ },
721
+ {
722
+ "match": "deepseek-v4-pro",
723
+ "ref": "deepseek-v4-pro"
724
+ },
725
+ {
726
+ "match": "kimi-k2.7-code",
727
+ "ref": "kimi-k2.7-code"
728
+ },
729
+ {
730
+ "match": "kimi-k3",
731
+ "ref": "kimi-k3"
732
+ },
733
+ {
734
+ "match": "kimi-k2.6",
735
+ "ref": "kimi-k2.6"
736
+ },
737
+ {
738
+ "match": "kimi",
739
+ "ref": "kimi-k2.5"
740
+ },
741
+ {
742
+ "match": "glm-4.5-airx",
743
+ "ref": "glm-4.5-airx"
744
+ },
745
+ {
746
+ "match": "glm-4.5-air",
747
+ "ref": "glm-4.5-air"
748
+ },
749
+ {
750
+ "match": "glm-4.5-x",
751
+ "ref": "glm-4.5-x"
752
+ },
753
+ {
754
+ "match": "glm-4.5-flash",
755
+ "ref": "glm-4.5-flash"
756
+ },
757
+ {
758
+ "match": "glm-4.5",
759
+ "ref": "glm-4.5"
760
+ },
761
+ {
762
+ "match": "glm-4.7-flashx",
763
+ "ref": "glm-4.7-flashx"
764
+ },
765
+ {
766
+ "match": "glm-4.7-flash",
767
+ "ref": "glm-4.7-flash"
768
+ },
769
+ {
770
+ "match": "glm-4.7",
771
+ "ref": "glm-4.7"
772
+ },
773
+ {
774
+ "match": "glm-4.6",
775
+ "ref": "glm-4.6"
776
+ },
777
+ {
778
+ "match": "glm-5.3-flash",
779
+ "ref": "glm-5.3-flash"
780
+ },
781
+ {
782
+ "match": "glm-5.3",
783
+ "ref": "glm-5.3"
784
+ },
785
+ {
786
+ "match": "glm-5-turbo",
787
+ "ref": "glm-5-turbo"
788
+ },
789
+ {
790
+ "match": "glm-5.2",
791
+ "ref": "glm-5.2"
792
+ },
793
+ {
794
+ "match": "glm-5.1",
795
+ "ref": "glm-5.1"
796
+ },
797
+ {
798
+ "match": "glm-5",
799
+ "ref": "glm-5"
800
+ },
801
+ {
802
+ "match": "fugu",
803
+ "ref": "sakana/fugu-ultra"
804
+ },
805
+ {
806
+ "match": "longcat",
807
+ "ref": "longcat-2.0"
808
+ },
809
+ {
810
+ "match": "step-3.7-flash",
811
+ "ref": "step-3.7-flash"
812
+ },
813
+ {
814
+ "match": "step-3.5-flash",
815
+ "ref": "step-3.5-flash"
816
+ },
817
+ {
818
+ "match": "stepfun",
819
+ "ref": "step-3.7-flash"
820
+ }
821
+ ]
822
+ }