Schema reference
Two records power the registry: Model and Provider.
Model records describe a specific API model. Provider records describe the source behind those models. Keep the shape stable; update values through review.
Model
One row per API-facing model id. This is what model pages, comparison, pricing tables, filters, and API clients consume.
| Field | Type | Required | Meaning |
|---|---|---|---|
id / object | uuid / "model" | Yes | Public UUID plus fixed object type. Detail and list responses both return these. |
provider | string | Yes | Stable provider slug, for example `openai`, `google`, `anthropic`, `qwen`. |
provider_name | string | No | Frontend-enriched provider display name when joined from provider records. |
icon_slug | string | No | Provider icon slug used by model cards, detail pages, and comparison tables. |
model | string | Yes | Stable API model id and URL slug. Use this for `/models/{slug}`. |
model_name | string | Yes | Human-readable display name. |
family / family_name / family_default | string / string / boolean | No | Model family grouping plus the default model marker for a family. |
version_tag / parent_id | string / uuid | No | Version label and parent model UUID when a model belongs to a versioned family. |
versions / is_latest | array / boolean | No | Version summaries returned on parent detail records, plus latest-version marker. |
status | enum | Yes | `active`, `preview`, `deprecated`, or `legacy`. |
is_flagship / is_recommended / is_featured / is_new | boolean | Yes | Discovery flags that drive sorting, badges, and homepage ranking. |
recommendation_badge | string | No | Optional badge text for curated recommendations. |
access_type | enum | Yes | `closed`, `open_weights`, or `open_source`. |
license / parameter_size | string | No | License name and model size such as `7B`, `72B`, or `405B` when published. |
tagline / description / best_for | string | No | Human-facing model copy for cards, detail pages, SEO, and use-case guidance. |
input_price / output_price | number | null | No | Detail response price per 1M tokens. Null means unknown or unpublished, not free. |
cached_input_price | number | null | No | Cached input price per 1M tokens when the provider publishes cache pricing. |
batch_input_price / batch_output_price | number | null | No | Batch API prices per 1M tokens when separate batch pricing exists. |
input_price_per_million / output_price_per_million | number | null | No | Lightweight list response pricing fields. Same unit as detail pricing. |
currency | string | Yes | Pricing currency. Current normalized public API value is usually `USD`. |
context_window / max_input_tokens / max_output_tokens | number | No | Token limits used by pricing tables, filters, and detail pages. |
knowledge_cutoff / training_data_info | string | No | Known training cutoff date and provider-published training data notes. |
input_modalities / output_modalities | string[] | No | Text, image, audio, video, embedding, score, or other supported modes. |
endpoints | object | Yes | Endpoint flags: chat_completions, responses, realtime, assistants, batch, embeddings, fine_tuning, image_generation, vision, speech_generation, transcription, translation, moderation, videos, image_edit. |
features | object | No | Capability flags: streaming, function_calling, structured_output, json_mode, distillation, reasoning, system_prompt. |
tools | object | No | Tool support flags: web_search, file_search, image_generation, code_interpreter, computer_use, mcp. |
config_parameters | array | No | Runtime parameter schema for UI controls. Each item contains name, use_template, localized label/help, type, required, default, min, max, precision, and options. |
rate_limits | array | No | Tiered limits with tier, rpm, tpm, rpd, and batch_limit. |
evaluation | object | No | Evaluation data: strengths, limitations, arena_elo, arena_rank, arena_ci, arena_votes, arena_category_ranks, official_benchmarks. |
use_cases | string[] | No | Product filters such as chat, coding, embedding, rerank, vision, audio. |
meta_title / meta_description | string | No | SEO title and description for public model detail pages. |
has_vision / has_audio / has_reasoning / has_multimodal / has_prediction | boolean | No | List response convenience flags derived from modalities, endpoints, and features. |
created_at / updated_at | number | omitted | No | Reserved client fields. Current public model API hides these timestamps. |
source_url / data_source / last_synced_at / sync_checksum | string | timestamp | Review only | Crawler and review provenance. Stored internally or in DataSync payloads, not public model response fields. |
Example JSON
Model record
{
"id": "8f4a2a66-7f1f-4d52-8b62-7f62a2c2a501",
"object": "model",
"provider": "qwen",
"provider_name": "Qwen",
"icon_slug": "qwen",
"model": "qwen3-max",
"model_name": "Qwen3-Max",
"family": "qwen3",
"family_name": "Qwen3",
"family_default": true,
"version_tag": "latest",
"parent_id": null,
"status": "active",
"is_flagship": true,
"is_recommended": true,
"is_featured": true,
"is_new": false,
"recommendation_badge": "Flagship",
"access_type": "closed",
"license": null,
"parameter_size": null,
"tagline": "Flagship Qwen model for reasoning and coding.",
"description": "A production model record with pricing, limits, capabilities, tools, and SEO fields.",
"best_for": "General chat, coding, reasoning, and agent workflows.",
"currency": "USD",
"input_price": 0.359,
"output_price": 1.434,
"cached_input_price": null,
"batch_input_price": null,
"batch_output_price": null,
"input_price_per_million": 0.359,
"output_price_per_million": 1.434,
"context_window": 262144,
"max_input_tokens": 262144,
"max_output_tokens": 32768,
"knowledge_cutoff": "2025-09-01",
"training_data_info": "Provider-published model documentation.",
"input_modalities": ["text", "image"],
"output_modalities": ["text"],
"endpoints": {
"chat_completions": true,
"responses": false,
"realtime": false,
"assistants": false,
"batch": true,
"embeddings": false,
"fine_tuning": false,
"image_generation": false,
"vision": true,
"speech_generation": false,
"transcription": false,
"translation": false,
"moderation": false,
"completions": false,
"videos": false,
"image_edit": false
},
"features": {
"streaming": true,
"function_calling": true,
"structured_output": true,
"json_mode": true,
"distillation": false,
"reasoning": true,
"system_prompt": true,
"logprobs": false,
"fine_tuning": false
},
"tools": {
"web_search": false,
"file_search": false,
"image_generation": false,
"code_interpreter": false,
"computer_use": false,
"mcp": true
},
"config_parameters": [
{
"name": "temperature",
"use_template": "temperature",
"label": { "en_US": "Temperature", "zh_Hans": "温度" },
"help": { "en_US": "Controls randomness.", "zh_Hans": "控制随机性。" },
"type": "float",
"required": false,
"default": 0.7,
"min": 0,
"max": 2,
"precision": 2,
"options": []
}
],
"rate_limits": [{ "tier": "default", "rpm": 60, "tpm": 120000, "rpd": 10000 }],
"evaluation": {
"strengths": ["reasoning", "coding"],
"limitations": ["Pricing varies by region"],
"arena_elo": null,
"arena_rank": null,
"arena_ci": null,
"arena_votes": null,
"arena_category_ranks": {},
"official_benchmarks": {}
},
"use_cases": ["text-chat", "reasoning", "coding"],
"meta_title": "Qwen3-Max model pricing, context window, and capabilities",
"meta_description": "Compare Qwen3-Max pricing, limits, modalities, and supported endpoints.",
"versions": [
{
"id": "qwen3-max",
"model_id": "qwen3-max",
"name": "Qwen3-Max",
"version_tag": "latest",
"release_date": "2025-09-01",
"status": "active",
"is_latest": true
}
],
"is_latest": true,
"has_vision": true,
"has_audio": false,
"has_reasoning": true,
"has_multimodal": true,
"has_prediction": false,
"created_at": null,
"updated_at": null,
"_review_payload_only": {
"source_url": "https://www.alibabacloud.com/help/en/model-studio/models",
"data_source": "crawler",
"last_synced_at": "2026-05-03T00:00:00Z",
"sync_checksum": "sha256:..."
}
}Provider
One row per model source. This controls provider pages, brand display, official links, and model grouping.
| Field | Type | Required | Meaning |
|---|---|---|---|
provider | string | Yes | Stable provider slug. This joins provider records to model records. |
provider_name | string | Yes | Display name. Example: Qwen should display as `Qwen`, not `Alibaba Cloud`. |
icon_slug / logo_url | string | No | Brand mark used in cards, nav, and provider pages. |
website | string | No | Canonical product or provider homepage. |
api_docs_url | string | No | Official API documentation link. |
pricing_url | string | No | Official pricing or billing page. |
tagline / description | string | No | Provider-level product context for SEO and provider pages. |
is_verified | boolean | Yes | Whether the provider profile has been reviewed. |
metadata.i18n | object | No | Localized provider copy when available. |
Example JSON
Provider record
{
"provider": "qwen",
"provider_name": "Qwen",
"website": "https://qwen.ai",
"api_docs_url": "https://help.aliyun.com/zh/model-studio/developer-reference/api-reference",
"pricing_url": "https://help.aliyun.com/zh/model-studio/developer-reference/tongyi-qianwen-metering-and-billing",
"tagline": "Qwen - Tongyi Qianwen model family",
"is_verified": true,
"metadata": { "i18n": { "zh": { "tagline": "Qwen - 通义千问模型家族" } } }
}