From 3481b4fd7ea358330c1589d5d0506a34a85ec501 Mon Sep 17 00:00:00 2001 From: Claude Date: Fri, 13 Nov 2026 03:24:21 +0000 Subject: [PATCH] docs: sync model catalog with dashboard API - gliner2.5-decide-v1: input $0.004 -> $0.03 (catalog table + card pill, text-classification table) - add d1.1 (Text Classification, decision, 65,536 max tokens, $0.03 input, output not billed): catalog row + card + task-library count 8 -> 9, text-classification table row and `## d1.1` section, model page, d1_1 playground spec, SystemOneRequest.model, docs.json nav, and the System One reference page (four models -> five) - audit script: check SystemOneRequest.model for decision models and the audio request enums for transcription/speech models, instead of the Responses/Chat enums they are not routable on d1.1 is routable only on POST /v1/systemone per the API description, and the payload ships no sample bodies for it, so its section documents the request only and no response example was invented. No models were removed; the API returns 31. Source: https://api-dashboard.zerogpu.ai/api/models Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01V6BqNyrYRBCR1fUq6WhACe --- .../skills/model-sync/scripts/audit-models.py | 26 + api-reference/models/d1-1.mdx | 19 + .../openapi/playgrounds/d1_1.openapi.json | 600 ++++++++++++++++++ api-reference/openapi/zerogpu.openapi.json | 2 +- api-reference/systemone.mdx | 7 +- docs.json | 1 + docs/model-catalog.mdx | 19 +- docs/text-classification.mdx | 51 +- 8 files changed, 717 insertions(+), 8 deletions(-) create mode 100644 api-reference/models/d1-1.mdx create mode 100644 api-reference/openapi/playgrounds/d1_1.openapi.json diff --git a/.claude/skills/model-sync/scripts/audit-models.py b/.claude/skills/model-sync/scripts/audit-models.py index 6473079..1b537d2 100755 --- a/.claude/skills/model-sync/scripts/audit-models.py +++ b/.claude/skills/model-sync/scripts/audit-models.py @@ -436,6 +436,10 @@ def model_prop(schema): chat_enum = model_prop("CreateChatCompletionRequest").get("enum", []) embedding_enum = model_prop("CreateEmbeddingRequest").get("enum", []) moderation_prop = json.dumps(model_prop("CreateModerationRequest")) + systemone_prop = json.dumps(model_prop("SystemOneRequest")) + transcription_enum = model_prop("CreateTranscriptionRequest").get("enum", []) + speech_enum = model_prop("CreateSpeechRequest").get("enum", []) + speech_multipart_enum = model_prop("CreateSpeechMultipartRequest").get("enum", []) print(f"API: {len(models)} models — {', '.join(api_ids)}") print(f"mode: {'FIX (table cells rewritten in place)' if args.fix else 'report only'}\n") @@ -511,9 +515,31 @@ def model_prop(schema): if task_api == "Text Moderation": if mid not in moderation_prop: notes.append(f"{oa}: not referenced by CreateModerationRequest.model") + elif m.get("taskType") == "decision": + # A decision model is routable only on /v1/systemone: it generates no + # text, so it belongs in SystemOneRequest.model, not in the + # Responses/Chat enums. Same shape as the moderation case above — + # SystemOneRequest.model carries no enum, so the whole property is + # searched. + if mid not in systemone_prop: + notes.append(f"{oa}: not referenced by SystemOneRequest.model") elif task_api == "Text Embedding": if mid not in embedding_enum: missing.append(f"{oa}: absent from the CreateEmbeddingRequest model enum") + elif m.get("taskType") == "transcription": + # Audio models are routable only on /v1/audio/*, never on Responses or + # Chat, so their own request schema is the one to check. + if mid not in transcription_enum: + missing.append( + f"{oa}: absent from the CreateTranscriptionRequest model enum" + ) + elif m.get("taskType") == "speech": + for schema, enum in ( + ("CreateSpeechRequest", speech_enum), + ("CreateSpeechMultipartRequest", speech_multipart_enum), + ): + if mid not in enum: + missing.append(f"{oa}: absent from the {schema} model enum") else: # The payload does not say which endpoints a model is routable on: # llama-3.1-8b-instruct-fast has sample_responses_body: null yet is fully diff --git a/api-reference/models/d1-1.mdx b/api-reference/models/d1-1.mdx new file mode 100644 index 0000000..da2db28 --- /dev/null +++ b/api-reference/models/d1-1.mdx @@ -0,0 +1,19 @@ +--- +title: "d1.1" +sidebarTitle: "d1.1" +description: "Model details for d1.1. LiquidAI's decision model, served on /v1/systemone with a 65,536-token limit." +openapi: "api-reference/openapi/playgrounds/d1_1.openapi.json POST /systemone" +--- + + + `d1.1` is routable only on `POST /v1/systemone`. It generates no text, so it is not + available on `/v1/responses` or `/v1/chat/completions`. + + +> LiquidAI's d1.1 is a decision model and the successor to d1. You send a state — text, JSON or images — together with several named questions, and every question is answered in a single forward pass: a `noul` question returns the probability of yes, a `choice` question picks one of your named options with a confidence and the full probabilities, and a `score` question rates the state on 2 to 10 ordered levels. No text is generated, so output tokens are always 0 and only input is billed. It is built for intent classification, agent routing, moderation, commercial-intent detection, triage and LLM-as-a-judge checks, and it does not chat, stream or call tools. + +## System One + +The request and response follow the [System One](/api-reference/systemone) decision schema. Unlike the ZLM classifiers, d1.1 has no taxonomy of its own: every question is answered against the criteria you supply. + +**References:** [Terms](https://www.liquid.ai/terms-conditions) • [Privacy](https://zerogpu.ai/privacy-policy) diff --git a/api-reference/openapi/playgrounds/d1_1.openapi.json b/api-reference/openapi/playgrounds/d1_1.openapi.json new file mode 100644 index 0000000..a82086d --- /dev/null +++ b/api-reference/openapi/playgrounds/d1_1.openapi.json @@ -0,0 +1,600 @@ +{ + "openapi": "3.1.0", + "info": { + "title": "d1.1 playground", + "version": "1.0", + "description": "Interactive playground for **d1.1**.\nModel is always `d1.1` on this page (shown in the form, not editable).\nd1.1 is a decision model, routable only on `/v1/systemone` — it generates no text and is not available on `/v1/responses` or `/v1/chat/completions`.\nSend the content as `state` and up to 16 named `questions`; each answer comes back under its question's name and `output_tokens` is always 0.\nAuthentication: `x-api-key`, or the OpenAI-compatible `Authorization: Bearer` header. `x-project-id` is optional." + }, + "servers": [ + { + "url": "https://api.zerogpu.ai/v1", + "description": "Production" + } + ], + "security": [ + { + "ApiKey": [] + }, + { + "BearerAuth": [] + } + ], + "paths": { + "/systemone": { + "post": { + "operationId": "createResponse_d1_1", + "summary": "Create a d1.1 decision", + "description": "Answer named `noul`, `choice` and `score` questions about `state`, compatible with TypeSafe System One.", + "tags": [ + "System One" + ], + "requestBody": { + "required": true, + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/SystemOneRequest" + }, + "example": { + "model": "d1.1", + "state": "The customer says their payment failed twice and they want someone to call them back today.", + "questions": { + "intent": { + "type": "choice", + "instructions": "Which intent best describes this message?", + "criteria": { + "payment_issue": "A problem paying or being charged", + "refund": "A request for money back", + "account_access": "Cannot sign in or recover an account", + "general_question": "Anything else" + } + }, + "handoff": { + "type": "noul", + "instructions": "Does this message need a human agent?" + }, + "urgency": { + "type": "score", + "instructions": "How urgent is this message?", + "criteria": [ + "None", + "Low", + "Medium", + "High" + ] + } + } + } + } + } + }, + "responses": { + "200": { + "description": "Success", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/SystemOneResponse" + } + } + } + }, + "400": { + "description": "The model does not support System One (`model_not_supported`)", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorResponse" + } + } + } + }, + "401": { + "description": "Unauthorized (missing API key)" + }, + "402": { + "description": "Insufficient quota (insufficient_quota)", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorResponse" + } + } + } + }, + "404": { + "description": "Model not found (`model_not_found`)", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorResponse" + } + } + } + }, + "422": { + "description": "Validation error", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/HTTPValidationError" + } + } + } + }, + "502": { + "description": "Upstream inference failed (`upstream_error`)", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorResponse" + } + } + } + } + } + } + } + }, + "components": { + "securitySchemes": { + "ApiKey": { + "type": "apiKey", + "in": "header", + "name": "x-api-key" + }, + "BearerAuth": { + "type": "http", + "scheme": "bearer", + "description": "OpenAI-compatible: send your ZeroGPU API key as `Authorization: Bearer `." + } + }, + "schemas": { + "ErrorResponse": { + "type": "object", + "additionalProperties": true + }, + "SystemOneRequest": { + "type": "object", + "required": [ + "model", + "state", + "questions" + ], + "properties": { + "model": { + "type": "string", + "const": "d1.1", + "default": "d1.1", + "example": "d1.1", + "description": "Model identifier (fixed for this playground)." + }, + "state": { + "description": "The content every question refers to: a string, object or array. An object or array is read as its JSON. For `zlm-v1-iab-domain-classifier` it must be a URL or domain.", + "anyOf": [ + { + "type": "string" + }, + { + "type": "object" + }, + { + "type": "array", + "items": {} + } + ] + }, + "questions": { + "type": "object", + "minProperties": 1, + "maxProperties": 16, + "description": "Named questions (up to 16). Keys are names you choose; each answer comes back under the same key.", + "additionalProperties": { + "$ref": "#/components/schemas/SystemOneQuestion" + } + } + } + }, + "SystemOneQuestion": { + "oneOf": [ + { + "$ref": "#/components/schemas/NoulQuestion" + }, + { + "$ref": "#/components/schemas/ChoiceQuestion" + }, + { + "$ref": "#/components/schemas/ScoreQuestion" + } + ], + "discriminator": { + "propertyName": "type", + "mapping": { + "noul": "#/components/schemas/NoulQuestion", + "choice": "#/components/schemas/ChoiceQuestion", + "score": "#/components/schemas/ScoreQuestion" + } + } + }, + "NoulQuestion": { + "type": "object", + "required": [ + "type" + ], + "description": "A yes/no (true/false) decision.", + "properties": { + "type": { + "type": "string", + "const": "noul" + }, + "instructions": { + "description": "The statement or question to evaluate.", + "anyOf": [ + { + "type": "string" + }, + { + "type": "object" + }, + { + "type": "array", + "items": {} + }, + { + "type": "null" + } + ] + }, + "criteria": { + "description": "Optional definitions of what counts as true and false.", + "anyOf": [ + { + "type": "object", + "properties": { + "true": { + "description": "Free-form text: a string, or an object or array that is read as its JSON.", + "anyOf": [ + { + "type": "string" + }, + { + "type": "object" + }, + { + "type": "array", + "items": {} + }, + { + "type": "null" + } + ] + }, + "false": { + "description": "Free-form text: a string, or an object or array that is read as its JSON.", + "anyOf": [ + { + "type": "string" + }, + { + "type": "object" + }, + { + "type": "array", + "items": {} + }, + { + "type": "null" + } + ] + } + } + }, + { + "type": "null" + } + ] + } + } + }, + "ChoiceQuestion": { + "type": "object", + "required": [ + "type", + "criteria" + ], + "description": "Pick one option.", + "properties": { + "type": { + "type": "string", + "const": "choice" + }, + "instructions": { + "description": "Free-form text: a string, or an object or array that is read as its JSON.", + "anyOf": [ + { + "type": "string" + }, + { + "type": "object" + }, + { + "type": "array", + "items": {} + }, + { + "type": "null" + } + ] + }, + "criteria": { + "type": "object", + "minProperties": 1, + "maxProperties": 32, + "description": "Choice names mapped to descriptions. A null description means the name alone is used.", + "additionalProperties": { + "description": "Free-form text: a string, or an object or array that is read as its JSON.", + "anyOf": [ + { + "type": "string" + }, + { + "type": "object" + }, + { + "type": "array", + "items": {} + }, + { + "type": "null" + } + ] + } + } + } + }, + "ScoreQuestion": { + "type": "object", + "required": [ + "type", + "criteria" + ], + "description": "Rate against an ordered rubric.", + "properties": { + "type": { + "type": "string", + "const": "score" + }, + "instructions": { + "description": "What to rate.", + "anyOf": [ + { + "type": "string" + }, + { + "type": "object" + }, + { + "type": "array", + "items": {} + }, + { + "type": "null" + } + ] + }, + "criteria": { + "type": "array", + "minItems": 1, + "maxItems": 32, + "description": "Ordered level descriptions. A level's position is its score, starting at 0.", + "items": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "object" + }, + { + "type": "array", + "items": {} + } + ] + } + } + } + }, + "SystemOneResponse": { + "type": "object", + "required": [ + "model", + "answers", + "usage" + ], + "properties": { + "model": { + "type": "string", + "description": "The model that answered." + }, + "answers": { + "type": "object", + "description": "Keyed by your question names, in request order.", + "additionalProperties": { + "$ref": "#/components/schemas/SystemOneAnswer" + } + }, + "usage": { + "$ref": "#/components/schemas/SystemOneUsage" + } + } + }, + "SystemOneAnswer": { + "oneOf": [ + { + "$ref": "#/components/schemas/NoulAnswer" + }, + { + "$ref": "#/components/schemas/ChoiceAnswer" + }, + { + "$ref": "#/components/schemas/ScoreAnswer" + } + ], + "discriminator": { + "propertyName": "type", + "mapping": { + "noul": "#/components/schemas/NoulAnswer", + "choice": "#/components/schemas/ChoiceAnswer", + "score": "#/components/schemas/ScoreAnswer" + } + } + }, + "NoulAnswer": { + "type": "object", + "required": [ + "type", + "noul" + ], + "properties": { + "type": { + "type": "string", + "const": "noul" + }, + "noul": { + "type": "number", + "minimum": 0, + "maximum": 1, + "description": "Probability of true. Near 1 favors true, near 0 favors false, near 0.5 is uncertain." + } + } + }, + "ChoiceAnswer": { + "type": "object", + "required": [ + "type", + "choice", + "confidence", + "probabilities" + ], + "properties": { + "type": { + "type": "string", + "const": "choice" + }, + "choice": { + "type": "string", + "description": "The most likely choice name." + }, + "confidence": { + "type": "number", + "minimum": 0, + "maximum": 1 + }, + "probabilities": { + "type": "object", + "description": "Keyed by choice name; sums to about 1.", + "additionalProperties": { + "type": "number" + } + } + } + }, + "ScoreAnswer": { + "type": "object", + "required": [ + "type", + "score", + "confidence", + "legend", + "probabilities" + ], + "properties": { + "type": { + "type": "string", + "const": "score" + }, + "score": { + "type": "number", + "description": "Probability-weighted average level; may fall between integers." + }, + "confidence": { + "type": "number", + "minimum": 0, + "maximum": 1 + }, + "legend": { + "type": "object", + "description": "Level key (\"0\", \"1\", ...) to its criteria text.", + "additionalProperties": {} + }, + "probabilities": { + "type": "object", + "description": "Keyed like `legend`; sums to about 1.", + "additionalProperties": { + "type": "number" + } + } + } + }, + "SystemOneUsage": { + "type": "object", + "required": [ + "input_tokens", + "output_tokens", + "total_tokens" + ], + "properties": { + "input_tokens": { + "type": "integer", + "description": "Tokens billed: the state plus the questions." + }, + "output_tokens": { + "type": "integer", + "description": "Always 0; System One output is not billed." + }, + "total_tokens": { + "type": "integer" + } + } + }, + "HTTPValidationError": { + "type": "object", + "description": "Request validation failure (System One's own error shape).", + "properties": { + "detail": { + "type": "array", + "items": { + "type": "object", + "required": [ + "loc", + "msg", + "type" + ], + "properties": { + "loc": { + "type": "array", + "items": { + "anyOf": [ + { + "type": "string" + }, + { + "type": "integer" + } + ] + } + }, + "msg": { + "type": "string" + }, + "type": { + "type": "string" + } + } + } + } + } + } + } + } +} diff --git a/api-reference/openapi/zerogpu.openapi.json b/api-reference/openapi/zerogpu.openapi.json index 61bd05c..f88e640 100644 --- a/api-reference/openapi/zerogpu.openapi.json +++ b/api-reference/openapi/zerogpu.openapi.json @@ -2321,7 +2321,7 @@ "properties": { "model": { "type": "string", - "description": "A model that supports System One: `zlm-v1-moderation-edge`, `zlm-v1-iab-classify-edge`, `zlm-v1-iab-domain-classifier` or `zlm-v2-iab-classify-edge-enriched`.", + "description": "A model that supports System One: `d1.1`, `zlm-v1-moderation-edge`, `zlm-v1-iab-classify-edge`, `zlm-v1-iab-domain-classifier` or `zlm-v2-iab-classify-edge-enriched`. `d1.1` is routable only on this endpoint.", "example": "zlm-v1-iab-classify-edge" }, "state": { diff --git a/api-reference/systemone.mdx b/api-reference/systemone.mdx index 1a0996c..70ad0ef 100644 --- a/api-reference/systemone.mdx +++ b/api-reference/systemone.mdx @@ -1,16 +1,17 @@ --- title: "System One" -description: "Answer yes/no, multiple-choice and rubric questions about any content with the ZLM classifiers." +description: "Answer yes/no, multiple-choice and rubric questions about any content with d1.1 and the ZLM classifiers." sidebarTitle: "System One" openapi: "api-reference/openapi/zerogpu.openapi.json POST /systemone" --- Send the content as `state` and ask up to 16 named `questions` about it in one call. Each answer comes back under the question's name, with calibrated probabilities. The request and response follow the [TypeSafe System One](https://api.typesafe.ai/openapi.json) decision schema, so an existing System One client works by changing the base URL and API key. -System One is available on four models, in addition to their existing endpoints: +System One is available on five models. For `d1.1` it is the only endpoint; the ZLM classifiers answer System One questions in addition to their existing endpoints: | Model | Also served on | Good for | | --- | --- | --- | +| [`d1.1`](/api-reference/models/d1-1) | — (System One only) | Any `noul`, `choice` or `score` question against criteria you supply: intent, routing, moderation, commercial intent, triage, LLM-as-a-judge. | | [`zlm-v1-moderation-edge`](/api-reference/models/zlm-v1-moderation-edge) | `/moderations` | Is this unsafe? Which moderation category applies? How severe is it? | | [`zlm-v1-iab-classify-edge`](/api-reference/models/zlm-v1-iab-classify-edge) | `/chat/completions`, `/responses` | Which IAB category matches? Does it belong to category X? How strongly? | | [`zlm-v1-iab-domain-classifier`](/api-reference/models/zlm-v1-iab-domain-classifier) | `/chat/completions`, `/responses` | The same questions about a domain. `state` must be a URL or domain. | @@ -28,6 +29,8 @@ Each model scores the content against its own labels: the 13 OpenAI moderation c A question whose criteria are not among the model's labels is answered by zero-shot classification instead. You can mix both kinds in one request. +`d1.1` has no taxonomy of its own. Every question is answered against the criteria in the request, in a single forward pass. + ## Limits and billing - Up to 16 questions per request, and up to 32 criteria per question. diff --git a/docs.json b/docs.json index 9b3b30c..236a394 100644 --- a/docs.json +++ b/docs.json @@ -117,6 +117,7 @@ "api-reference/models/gliner2-5-base-v1", "api-reference/models/gliner2-5-multi-v1", "api-reference/models/gliner2-5-decide-v1", + "api-reference/models/d1-1", "api-reference/models/lfm2-5-1-2b-instruct", "api-reference/models/lfm2-5-1-2b-thinking", "api-reference/models/llama-3-1-8b-instruct-fast", diff --git a/docs/model-catalog.mdx b/docs/model-catalog.mdx index 2a69216..ad618f2 100644 --- a/docs/model-catalog.mdx +++ b/docs/model-catalog.mdx @@ -11,6 +11,7 @@ Sections on this page: [At a glance](#at-a-glance) (pricing table), [Detailed mo | Model | Input /1M | Output /1M | Cached input /1M | Task | Max tokens | | --- | ---: | ---: | ---: | --- | ---: | +| d1.1 d1.1 | $0.03 | Not billed | — | Text Classification | 65,536 | | deepseek-v4.1-flash deepseek-v4.1-flash | $0.30 | $1.20 | $0.003 | Text Generation | 1,048,576 | | glm-5.3-flash glm-5.3-flash | $0.10 | $0.35 | $0.015 | Text Generation | 1,048,576 | | gpt-5.4-nano gpt-5.4-nano | $0.20 | $1.25 | $0.02 | Text Generation | 400,000 | @@ -35,12 +36,14 @@ Sections on this page: [At a glance](#at-a-glance) (pricing table), [Detailed mo | zlm-v1-tool-decisions | $0.05 | $0.15 | — | Text Generation | 24,576 | | unlimited-ocr unlimited-ocr | Per page | Per page | — | Text Generation | — | | zlm-v1-signal-extract zlm-v1-signal-extract | $0.04 | $0.10 | — | Text Classification | 400 | -| gliner2.5-decide-v1 gliner2.5-decide-v1 | $0.004 | Not billed | — | Text Classification | — | +| gliner2.5-decide-v1 gliner2.5-decide-v1 | $0.03 | Not billed | — | Text Classification | — | | all-minilm-l6-v2 all-minilm-l6-v2 | $0.004 | Not billed | — | Text Embedding | 512 | | bge-small-en-v1.5 bge-small-en-v1.5 | $0.004 | Not billed | — | Text Embedding | 512 | `unlimited-ocr` is billed per page of document rather than per token, so its per-token columns read `Per page`. It is called on `/v1/responses` and `/v1/chat/completions` like the other text models. +`d1.1` is a decision model: it is called on `POST /v1/systemone` rather than `/v1/responses`, generates no text, and bills input tokens only, so its **Output /1M** column reads `Not billed`. See [System One](/api-reference/systemone). + [Audio models](/docs/audio) are called on `POST /v1/audio/transcriptions` and `POST /v1/audio/speech` rather than `/v1/responses`, and are billed per minute of audio instead of per token, rounded up to the next whole minute with a one-minute minimum. `chatterbox` is metered at the same per-minute rate but rounded up to the next whole second of generated audio, per request. | Model | Price /min | Task | Limits | @@ -52,6 +55,16 @@ Sections on this page: [At a glance](#at-a-glance) (pricing table), [Detailed mo ## Detailed model cards + +
+ d1.1 +
d1.1
+
+ +
65,536 max tokens\$0.03 / 1M input
+ + LiquidAI's decision model and the successor to d1. Send a state — text, JSON or images — together with several named questions to `/v1/systemone`, and every question is answered in a single forward pass… +
deepseek-v4.1-flash @@ -287,7 +300,7 @@ Sections on this page: [At a glance](#at-a-glance) (pricing table), [Detailed mo
gliner2.5-decide-v1
-
486M params\$0.004 / 1M input
+
486M params\$0.03 / 1M input
Fastino's GLiNER2.5 Decide is a 486M-parameter zero-shot decision and classification model built for operational AI workflows. You supply the labels or decision options at request time to classify intent, route agents and tools, triage support requests, make moderation decisions…
@@ -366,7 +379,7 @@ Sections on this page: [At a glance](#at-a-glance) (pricing table), [Detailed mo 3 models: `whisper-tiny`, `chatterbox-nano`, `chatterbox` - 8 models: `zlm-v2-iab-classify-edge-enriched`, `zlm-v1-iab-classify-edge`, `zlm-v1-iab-domain-classifier`, `deberta-v3-small`, `zlm-v1-signal-extract`, `gliner2.5-base-v1`, `gliner2.5-multi-v1`, `gliner2.5-decide-v1` + 9 models: `d1.1`, `zlm-v2-iab-classify-edge-enriched`, `zlm-v1-iab-classify-edge`, `zlm-v1-iab-domain-classifier`, `deberta-v3-small`, `zlm-v1-signal-extract`, `gliner2.5-base-v1`, `gliner2.5-multi-v1`, `gliner2.5-decide-v1` 1 model: `zlm-v1-moderation-edge` diff --git a/docs/text-classification.mdx b/docs/text-classification.mdx index 89790d1..0a99dda 100644 --- a/docs/text-classification.mdx +++ b/docs/text-classification.mdx @@ -3,10 +3,11 @@ title: "Text Classification" description: "Edge models that map text to labels, IAB taxonomy, and audience signals." --- -Text classification models map a passage to predefined labels or categories. ZeroGPU offers IAB taxonomy classifiers tuned for ad tech, a general zero-shot classifier where you supply the candidate labels, and the schema-based GLiNER2.5 models, which classify, extract entities, fill a JSON schema, and make dynamic-label decisions such as intent and routing from the same request. Each request is shown for both the [Responses API](/api-reference/responses) and the OpenAI-compatible [Chat Completions API](/api-reference/chat-completions). +Text classification models map a passage to predefined labels or categories. ZeroGPU offers IAB taxonomy classifiers tuned for ad tech, a general zero-shot classifier where you supply the candidate labels, and the schema-based GLiNER2.5 models, which classify, extract entities, fill a JSON schema, and make dynamic-label decisions such as intent and routing from the same request. Each request is shown for both the [Responses API](/api-reference/responses) and the OpenAI-compatible [Chat Completions API](/api-reference/chat-completions) — except `d1.1`, the decision model, which is called on [`POST /v1/systemone`](/api-reference/systemone) and neither chats nor streams. | Model | Input /1M | Output /1M | Max tokens | | --- | ---: | ---: | ---: | +| d1.1 d1.1 | $0.03 | Not billed | 65,536 | | zlm-v1-iab-classify-edge zlm-v1-iab-classify-edge | $0.02 | $0.05 | 400 | | zlm-v2-iab-classify-edge-enriched zlm-v2-iab-classify-edge-enriched | $0.05 | $0.15 | 800 | | zlm-v1-iab-domain-classifier zlm-v1-iab-domain-classifier | $0.02 | $0.05 | 100 | @@ -14,7 +15,53 @@ Text classification models map a passage to predefined labels or categories. Zer | zlm-v1-signal-extract zlm-v1-signal-extract | $0.04 | $0.10 | 400 | | gliner2.5-base-v1 gliner2.5-base-v1 | $0.02 | $0.05 | — | | gliner2.5-multi-v1 gliner2.5-multi-v1 | $0.02 | $0.05 | — | -| gliner2.5-decide-v1 gliner2.5-decide-v1 | $0.004 | Not billed | — | +| gliner2.5-decide-v1 gliner2.5-decide-v1 | $0.03 | Not billed | — | + +## d1.1 + + + `d1.1` is routable only on `POST /v1/systemone`. It generates no text, so it is not + available on `/v1/responses` or `/v1/chat/completions`. + + +> LiquidAI's d1.1 is a decision model and the successor to d1. You send a state — text, JSON or images — together with several named questions, and every question is answered in a single forward pass: a `noul` question returns the probability of yes, a `choice` question picks one of your named options with a confidence and the full probabilities, and a `score` question rates the state on 2 to 10 ordered levels. No text is generated, so output tokens are always 0 and only input is billed. It is built for intent classification, agent routing, moderation, commercial-intent detection, triage and LLM-as-a-judge checks, and it does not chat, stream or call tools. + +**References:** [Terms](https://www.liquid.ai/terms-conditions) • [Privacy](https://zerogpu.ai/privacy-policy) + +Send the content as `state` and the questions you want answered as a named `questions` map. The [System One reference](/api-reference/systemone) documents the three question types and the shape of each answer. + +```bash System One +curl https://api.zerogpu.ai/v1/systemone \ + -H "content-type: application/json" \ + -H "x-api-key: YOUR_API_KEY" \ + -d '{ + "model": "d1.1", + "state": "The customer says their payment failed twice and they want someone to call them back today.", + "questions": { + "intent": { + "type": "choice", + "instructions": "Which intent best describes this message?", + "criteria": { + "payment_issue": "A problem paying or being charged", + "refund": "A request for money back", + "account_access": "Cannot sign in or recover an account", + "general_question": "Anything else" + } + }, + "handoff": { + "type": "noul", + "instructions": "Does this message need a human agent?" + }, + "urgency": { + "type": "score", + "instructions": "How urgent is this message?", + "criteria": ["None", "Low", "Medium", "High"] + } + } +}' +``` + +Each answer comes back under the name you gave its question, alongside a `usage` block in which `output_tokens` is always 0. ## zlm-v1-iab-classify-edge