@grida/ai-models 2.20261115.0 → 2.20261115.1

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.
package/PROVENANCE.md CHANGED
@@ -1910,3 +1910,65 @@ unspecified; shared effort names do not promise a particular native thinking
1910
1910
  level. Sampling controls, explicit caches, grounding, file/audio/video input,
1911
1911
  background generation and continuation are outside this service contract.
1912
1912
  Native model facts and other provider bindings remain unchanged.
1913
+
1914
+ ## Gemini Decision 1 — 2026-11-15
1915
+
1916
+ Google's [changelog](https://aistudio.google.com/docs/changelog) dates the first
1917
+ public release of `gemini-decision-1` and the Decisions API to 2026-11-11. The
1918
+ [model page](https://aistudio.google.com/docs/models/gemini-decision-1) describes
1919
+ a specialized multimodal model, without naming an underlying model. The card
1920
+ therefore omits `underlying_model`; no Gemini chat parent or self-reference is
1921
+ inferred. It adds facts under the existing schema 2 contract, with no new shared
1922
+ execution specification.
1923
+
1924
+ The [native guide](https://aistudio.google.com/docs/decisions-api) uses `input`
1925
+ as a string, `Content` or `Content[]`, and named question/answer maps with
1926
+ `noul`, `choice` and `score`. The 262,144-token context includes shared input
1927
+ and all questions; it permits 1–1,000 questions, 2–255 choice options and 2–255
1928
+ score levels. The native content blocks include text, images, audio, video and
1929
+ PDF. Raw base64 data, Files API URIs, native resolution values
1930
+ `low`/`medium`/`high`/`ultra_high`, MIME restrictions and media size limits remain
1931
+ qualified rather than translated into OpenAI image fields or a new enum. Only
1932
+ the representable string input form is listed; this is not a text-only claim.
1933
+ The guide defers media counts to the Interactions API, so no guessed numeric
1934
+ image cap is recorded. Safety blocking is request-level HTTP 400, not a sourced
1935
+ partial-refusal guarantee. Confidence semantics remain native to each interface;
1936
+ applications must choose thresholds using their own labeled examples.
1937
+
1938
+ Google's [Standard paid pricing](https://aistudio.google.com/docs/pricing)
1939
+ charges $0.10/M input tokens across supported modalities. Shared input is
1940
+ counted once, and each question contributes instructions, criteria and formatting
1941
+ tokens. The source labels output pricing inapplicable because output usage is
1942
+ zero, and caching unavailable. The native binding records its input rate only;
1943
+ inapplicable output/cache rates are not invented. Batch, Flex and Priority are
1944
+ unsupported; [project-tier quotas](https://aistudio.google.com/docs/rate-limits)
1945
+ are not intrinsic model limits.
1946
+
1947
+ [Vercel's model-specific announcement](https://vercel.com/changelog/gemini-decision-1-now-available-on-ai-gateway)
1948
+ confirms its Decisions and TypeSafe-compatible text/image interfaces and additional
1949
+ PDF/audio/video input through AI SDK 7. The card's listed `decisions` endpoint
1950
+ records the [OpenAI-compatible facade](https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-decisions):
1951
+ question arrays, `predicate`, user messages, inline image data URLs, 1–255 choices
1952
+ and 2–10 score levels. Its published refusal shape is not a guarantee that native
1953
+ Google safety errors become per-question refusals. The model page's AI SDK
1954
+ `state`/named-question/`boolean` example is another client interface, not that
1955
+ facade's REST request. The [endpoint feed](https://ai-gateway.vercel.sh/v1/models/google/gemini-decision-1/endpoints)
1956
+ independently publishes context 262144 and regular input/completion rates of
1957
+ $0.10/$0 per million, with discount zero and implicit caching disabled. Generic
1958
+ reasoning/sampling/max-token and generation/tool price metadata is not copied
1959
+ into the decision contract.
1960
+
1961
+ [OpenRouter's model page](https://openrouter.ai/google/gemini-decision-1) and
1962
+ [Google AI Studio endpoint](https://openrouter.ai/api/v1/models/google/gemini-decision-1/endpoints)
1963
+ now establish availability, despite the general model feed omitting this alpha
1964
+ model. Its independent published regular input/completion rates are $0.10/$0
1965
+ per million with discount zero; context is 262144. Its
1966
+ [alpha API](https://openrouter.ai/docs/api/api-reference/alphadecisions/submit-a-decisions-request)
1967
+ uses `state` and named questions, distinct from Google's `input` and OpenAI's
1968
+ array envelope. The [multimodal guide](https://openrouter.ai/docs/guides/community/multimodal-decisions)
1969
+ explicitly describes its Google adapter's top-level image, audio, video and file
1970
+ parts, including public HTTPS media URLs and PDF/CSV/Office/OpenDocument files.
1971
+ Image detail `low`/`high` is forwarded and `auto` is dropped. Unpublished media
1972
+ counts, cache prices, question counts and context bands remain omitted. Generic
1973
+ tool-choice, web-search and positive max-completion metadata do not establish
1974
+ those features or charges for this decision offering.
package/README.md CHANGED
@@ -5,7 +5,7 @@ provider bindings, and published prices, with the labs that make them and their
5
5
  logos. Every fact cites a first-party source.
6
6
 
7
7
  This package is the data, published unchanged from release
8
- [`v2.20261115.0`](https://github.com/gridaco/models/releases/tag/v2.20261115.0) of [gridaco/models](https://github.com/gridaco/models). It contains no code.
8
+ [`v2.20261115.1`](https://github.com/gridaco/models/releases/tag/v2.20261115.1) of [gridaco/models](https://github.com/gridaco/models). It contains no code.
9
9
 
10
10
  | file | contents |
11
11
  | --- | --- |
@@ -26,7 +26,7 @@ From a CDN, without installing:
26
26
 
27
27
  ```
28
28
  https://cdn.jsdelivr.net/npm/@grida/ai-models@2/models.json
29
- https://cdn.jsdelivr.net/npm/@grida/ai-models@2.20261115.0/models.json
29
+ https://cdn.jsdelivr.net/npm/@grida/ai-models@2.20261115.1/models.json
30
30
  ```
31
31
 
32
32
  ## Versions
package/manifest.json CHANGED
@@ -1,19 +1,19 @@
1
1
  {
2
2
  "name": "grida-models",
3
- "version": "v2.20261115.0",
3
+ "version": "v2.20261115.1",
4
4
  "schema_version": 2,
5
- "commit": "26cd2de507d29961546dabe66d83fbb2063d0bc0",
6
- "created": "2026-11-15T10:57:18Z",
7
- "source": "https://github.com/gridaco/models/tree/26cd2de507d29961546dabe66d83fbb2063d0bc0/catalogue",
8
- "digest": "sha256:cf33026409454b86b5da940a849f41f6a1509dd76ba1918ee5b29210eaf181d0",
5
+ "commit": "b74e3e629417ee6508d61bb09d0baf13c4f8c86d",
6
+ "created": "2026-11-15T12:15:28Z",
7
+ "source": "https://github.com/gridaco/models/tree/b74e3e629417ee6508d61bb09d0baf13c4f8c86d/catalogue",
8
+ "digest": "sha256:0f65b969148ce68ecc50d56c479c783d689071f4d609a066de7941a078ef75bd",
9
9
  "counts": {
10
- "cards": 106,
10
+ "cards": 107,
11
11
  "labs": 18
12
12
  },
13
13
  "files": {
14
14
  "PROVENANCE.md": {
15
- "sha256": "b691da0cb87742ff9667a9bb6362f59655c24e14677555817a7e30b9c14c0fd8",
16
- "bytes": 102548
15
+ "sha256": "8b2924b4800f4af6f7f37041e045ae4ae3abdcb60dda6dc9e80e37f2af48d69e",
16
+ "bytes": 106896
17
17
  },
18
18
  "README.md": {
19
19
  "sha256": "02a012d4a62236c014e5f655e31d4e934981520788bde96788d1db3ba3eb9322",
@@ -96,8 +96,8 @@
96
96
  "bytes": 330
97
97
  },
98
98
  "models.json": {
99
- "sha256": "c01b4ed634e45e5e45b94ec05fa957f73e99e2f866b7028ff79217c29dd692ec",
100
- "bytes": 1123858
99
+ "sha256": "9ac4dfd712432a921d618da9544e7b22733ea8f7224ac48b3da8ef9d2388962a",
100
+ "bytes": 1136190
101
101
  },
102
102
  "schemas/labs.schema.json": {
103
103
  "sha256": "a720f50f9ee5c3ddd2fccf50e38ddecec686896b768a52cc987df52668171f07",
package/models.json CHANGED
@@ -28109,6 +28109,241 @@
28109
28109
  }
28110
28110
  },
28111
28111
  "decision.models": {
28112
+ "google/gemini-decision-1": {
28113
+ "id": "google/gemini-decision-1",
28114
+ "vendor": "google",
28115
+ "label": "Gemini Decision 1",
28116
+ "reference_provider": "google",
28117
+ "release": {
28118
+ "date": "2026-11-11",
28119
+ "basis": "model",
28120
+ "source_url": "https://aistudio.google.com/docs/changelog"
28121
+ },
28122
+ "providers": {
28123
+ "google": {
28124
+ "provider": "google",
28125
+ "id": "gemini-decision-1",
28126
+ "url": "https://aistudio.google.com/docs/models/gemini-decision-1",
28127
+ "source_url": "https://aistudio.google.com/docs/decisions-api",
28128
+ "qualifications": [
28129
+ {
28130
+ "source_url": "https://aistudio.google.com/docs/models/gemini-decision-1",
28131
+ "notes": "Google describes a specialized multimodal model for the Decisions API without naming an underlying model. No parent relationship or inherited chat capabilities are recorded."
28132
+ },
28133
+ {
28134
+ "source_url": "https://aistudio.google.com/docs/decisions-api",
28135
+ "notes": "The native v1beta Decisions API accepts input as a string, Content or Content[], with text, image, audio, video and PDF blocks, and questions/answers keyed by name. The structured input and Google's low/medium/high/ultra_high resolution vocabulary are not fully represented by decision_contract's existing shorthand. Raw inline base64, Files API URIs and public or signed HTTPS URLs are distinct native transports; no OpenAI image-form list is substituted."
28136
+ },
28137
+ {
28138
+ "source_url": "https://aistudio.google.com/docs/pricing",
28139
+ "notes": "Standard paid input is USD 0.10 per million tokens across all supported modalities. Shared input is billed once; question instructions, criteria and formatting add input tokens. Output pricing is not applicable because output usage is zero; context caching is unavailable. Inapplicable output/cache rates are omitted, rather than recorded as free generation or caching."
28140
+ },
28141
+ {
28142
+ "source_url": "https://aistudio.google.com/docs/decisions-api",
28143
+ "notes": "Questions require text instructions. Choice criteria use string keys; score criteria are ordered rubric levels. Confidence and probability semantics belong to this native interface and require thresholds calibrated on the consumer's own examples. Blocked input returns a request-level HTTP 400; no per-question refusal contract is recorded."
28144
+ }
28145
+ ],
28146
+ "token_pricing": {
28147
+ "source_url": "https://aistudio.google.com/docs/pricing",
28148
+ "unit": "usd_per_token",
28149
+ "scope": "Standard paid Decisions input across text, image, video, audio and PDF; shared input counted once plus question tokens. Output pricing is inapplicable and context caching is unavailable.",
28150
+ "input": "0.0000001"
28151
+ },
28152
+ "capabilities": {
28153
+ "source_url": "https://aistudio.google.com/docs/models/gemini-decision-1",
28154
+ "image_input": true
28155
+ },
28156
+ "limits": {
28157
+ "source_url": "https://aistudio.google.com/docs/models/gemini-decision-1",
28158
+ "context_tokens": 262144
28159
+ },
28160
+ "decision_contract": {
28161
+ "source_url": "https://aistudio.google.com/docs/decisions-api",
28162
+ "question_types": [
28163
+ "noul",
28164
+ "choice",
28165
+ "score"
28166
+ ],
28167
+ "input_forms": [
28168
+ "string"
28169
+ ],
28170
+ "question_container": "named_object",
28171
+ "maximum_questions": 1000,
28172
+ "minimum_choices": 2,
28173
+ "maximum_choices": 255,
28174
+ "minimum_score_levels": 2,
28175
+ "maximum_score_levels": 255
28176
+ },
28177
+ "endpoints": [
28178
+ {
28179
+ "id": "gemini-decision-1",
28180
+ "input": "text",
28181
+ "interface": "gemini_decisions",
28182
+ "source_url": "https://aistudio.google.com/docs/decisions-api"
28183
+ }
28184
+ ]
28185
+ },
28186
+ "vercel": {
28187
+ "provider": "vercel",
28188
+ "id": "google/gemini-decision-1",
28189
+ "url": "https://vercel.com/ai-gateway/models/gemini-decision-1",
28190
+ "source_url": "https://vercel.com/changelog/gemini-decision-1-now-available-on-ai-gateway",
28191
+ "qualifications": [
28192
+ {
28193
+ "source_url": "https://vercel.com/changelog/gemini-decision-1-now-available-on-ai-gateway",
28194
+ "notes": "Gemini Decision 1 is available through Vercel's Decisions API, TypeSafe-compatible API and AI SDK. The listed decisions endpoint records the OpenAI-compatible text/image facade; PDF/audio/video input is documented through AI SDK 7 and is not claimed for that endpoint."
28195
+ },
28196
+ {
28197
+ "source_url": "https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-decisions",
28198
+ "notes": "The compatible facade uses predicate/choice/score question arrays and string or user-message input, rather than Google's native noul/named-map/Content envelope. It requires inline image data URLs, accepts one to 255 choices and two to ten score levels, and rejects boolean/string spelling collisions. The source documents per-question refusal positions and input billing; this does not promise that native Google safety errors become partial refusals."
28199
+ },
28200
+ {
28201
+ "source_url": "https://ai-gateway.vercel.sh/v1/models/google/gemini-decision-1/endpoints",
28202
+ "notes": "The Google endpoint publishes a 262144-token context, input USD 0.0000001/token, zero completion cost and no implicit caching, with discount zero. Generic reasoning, sampling and max_tokens metadata is not a Decisions capability contract; native chat controls and extra zero-priced generation/tool meters are not inherited."
28203
+ }
28204
+ ],
28205
+ "token_pricing": {
28206
+ "source_url": "https://ai-gateway.vercel.sh/v1/models/google/gemini-decision-1/endpoints",
28207
+ "unit": "usd_per_token",
28208
+ "scope": "Published decision endpoint regular base input/completion rates; no unrecorded context bands, regional modifiers or cache rates are inferred.",
28209
+ "input": "0.0000001",
28210
+ "output": "0"
28211
+ },
28212
+ "capabilities": {
28213
+ "source_url": "https://vercel.com/changelog/gemini-decision-1-now-available-on-ai-gateway",
28214
+ "image_input": true
28215
+ },
28216
+ "limits": {
28217
+ "source_url": "https://ai-gateway.vercel.sh/v1/models/google/gemini-decision-1/endpoints",
28218
+ "context_tokens": 262144
28219
+ },
28220
+ "decision_contract": {
28221
+ "source_url": "https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-decisions",
28222
+ "question_types": [
28223
+ "predicate",
28224
+ "choice",
28225
+ "score"
28226
+ ],
28227
+ "input_forms": [
28228
+ "string",
28229
+ "user_messages"
28230
+ ],
28231
+ "question_container": "array",
28232
+ "image_input_forms": [
28233
+ "data_url"
28234
+ ],
28235
+ "image_details": [
28236
+ "low",
28237
+ "high",
28238
+ "auto",
28239
+ "original"
28240
+ ],
28241
+ "maximum_images": 128,
28242
+ "minimum_choices": 1,
28243
+ "maximum_choices": 255,
28244
+ "minimum_score_levels": 2,
28245
+ "maximum_score_levels": 10,
28246
+ "typed_choice_values": false
28247
+ },
28248
+ "endpoints": [
28249
+ {
28250
+ "id": "google/gemini-decision-1",
28251
+ "input": "text",
28252
+ "interface": "decisions",
28253
+ "source_url": "https://vercel.com/docs/ai-gateway/sdks-and-apis/openai-decisions"
28254
+ }
28255
+ ]
28256
+ },
28257
+ "openrouter": {
28258
+ "provider": "openrouter",
28259
+ "id": "google/gemini-decision-1",
28260
+ "url": "https://openrouter.ai/google/gemini-decision-1",
28261
+ "source_url": "https://openrouter.ai/google/gemini-decision-1",
28262
+ "qualifications": [
28263
+ {
28264
+ "source_url": "https://openrouter.ai/docs/api/api-reference/alphadecisions/submit-a-decisions-request",
28265
+ "notes": "The alpha Decisions API uses state, named questions, criteria and noul/choice/score, with answers keyed by name. Its envelope is distinct from Google's input field and OpenAI's question arrays. The alpha shape may change; a listing does not establish native Google or OpenAI SDK compatibility."
28266
+ },
28267
+ {
28268
+ "source_url": "https://openrouter.ai/docs/guides/community/multimodal-decisions",
28269
+ "notes": "The Google adapter reads image_url, input_audio, video_url and file parts in the top-level state array. Gemini accepts inline image data URLs and public HTTPS image/video URLs; file parts additionally accept PDF, CSV, Office and OpenDocument types under the provider's extension and MIME rules. low/high image detail is forwarded as resolution and auto is dropped. No shared media-count limit is published for this adapter; native-provider limits still apply."
28270
+ },
28271
+ {
28272
+ "source_url": "https://openrouter.ai/api/v1/models/google/gemini-decision-1/endpoints",
28273
+ "notes": "The model page and Google AI Studio endpoint establish availability even though the general /api/v1/models feed does not list this alpha model. The endpoint publishes context 262144, regular input USD 0.0000001/token and completion zero with discount zero. Generic tool-choice, web-search price and positive max-completion metadata do not establish native Decisions tools, output generation or additional charges; cache rates and unrecorded bands remain omitted."
28274
+ }
28275
+ ],
28276
+ "token_pricing": {
28277
+ "source_url": "https://openrouter.ai/api/v1/models/google/gemini-decision-1/endpoints",
28278
+ "unit": "usd_per_token",
28279
+ "scope": "Published regular base decision input/completion rates; cache prices and unrecorded context/region modifiers are not inferred.",
28280
+ "input": "0.0000001",
28281
+ "output": "0"
28282
+ },
28283
+ "capabilities": {
28284
+ "source_url": "https://openrouter.ai/docs/guides/community/multimodal-decisions",
28285
+ "image_input": true
28286
+ },
28287
+ "limits": {
28288
+ "source_url": "https://openrouter.ai/api/v1/models/google/gemini-decision-1/endpoints",
28289
+ "context_tokens": 262144
28290
+ },
28291
+ "decision_contract": {
28292
+ "source_url": "https://openrouter.ai/docs/guides/community/multimodal-decisions",
28293
+ "question_types": [
28294
+ "noul",
28295
+ "choice",
28296
+ "score"
28297
+ ],
28298
+ "input_forms": [
28299
+ "state"
28300
+ ],
28301
+ "question_container": "named_object",
28302
+ "image_input_forms": [
28303
+ "data_url",
28304
+ "http_url"
28305
+ ],
28306
+ "image_details": [
28307
+ "auto",
28308
+ "low",
28309
+ "high"
28310
+ ]
28311
+ },
28312
+ "endpoints": [
28313
+ {
28314
+ "id": "google/gemini-decision-1",
28315
+ "input": "text",
28316
+ "interface": "openrouter_alpha_decisions",
28317
+ "source_url": "https://openrouter.ai/docs/api/api-reference/alphadecisions/submit-a-decisions-request",
28318
+ "decision_contract": {
28319
+ "source_url": "https://openrouter.ai/google/gemini-decision-1",
28320
+ "maximum_choices": 255,
28321
+ "maximum_score_levels": 255
28322
+ }
28323
+ }
28324
+ ]
28325
+ }
28326
+ },
28327
+ "provider_coverage": {
28328
+ "google": {
28329
+ "disposition": "listed",
28330
+ "source_url": "https://aistudio.google.com/docs/models/gemini-decision-1"
28331
+ },
28332
+ "vercel": {
28333
+ "disposition": "listed",
28334
+ "source_url": "https://vercel.com/changelog/gemini-decision-1-now-available-on-ai-gateway"
28335
+ },
28336
+ "openrouter": {
28337
+ "disposition": "listed",
28338
+ "source_url": "https://openrouter.ai/google/gemini-decision-1"
28339
+ },
28340
+ "fal": {
28341
+ "disposition": "excluded",
28342
+ "source_url": "https://fal.ai/models",
28343
+ "reason": "Catalogue coverage includes fal for media models only; no decision API compatibility is claimed."
28344
+ }
28345
+ }
28346
+ },
28112
28347
  "openai/gpt-6-luna-decisions": {
28113
28348
  "id": "openai/gpt-6-luna-decisions",
28114
28349
  "vendor": "openai",
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@grida/ai-models",
3
- "version": "2.20261115.0",
3
+ "version": "2.20261115.1",
4
4
  "description": "A hand-curated, source-backed catalogue of AI models, their labs, and published prices.",
5
5
  "keywords": [
6
6
  "ai",