> ## Documentation Index > Fetch the complete documentation index at: https://docs.routerlink.ai/llms.txt > Use this file to discover all available pages before exploring further. # Models > Discover and use the current RouterLink model catalog across supported vendors and provider channels through a unified API. # Models RouterLink exposes the current model catalog from its live registry. The catalog size changes as models and provider channels are added, updated, or deprecated, so use the `total` field from `GET /api/v1/models` as the source of truth instead of hardcoding a model count. Browse the full catalog at [**https://routerlink.ai/models**](https://routerlink.ai/models). *** ## API Reference ### List Models Retrieves a paginated list of logical models with metadata, capability tags, provider-channel aliases, and context/output limits. ```http theme={null} GET /api/v1/models ``` #### Parameters | Parameter | Type | Default | Description | | ----------------- | ---------- | ------- | ------------------------------------------------------------------------------------------ | | `page` | `int` | `1` | Page number | | `page_size` | `int` | `20` | Items per page (max `100`) | | `model_name` | `string` | - | Case-insensitive fuzzy search on the model display name | | `provider` | `string` | - | Filter by a single provider-channel alias (backward compatible) | | `providers` | `string[]` | - | Multi-select provider-channel filter (max 20 values) | | `vendor` | `string` | - | Filter by a single vendor/model owner (backward compatible) | | `vendors` | `string[]` | - | Multi-select vendor filter (max 20 values) | | `label` | `string` | - | Filter by label: `Recommended`, `Flagship`, `Free for Now`, or `Deprecated` | | `tags` | `string[]` | - | Capability/modality filter (max 10 values), e.g. `Text`, `Image`, `File`, `Audio`, `Video` | | `show_deprecated` | `bool` | `true` | Include deprecated models in results | | `context_min` | `int` | - | Minimum context window length in tokens | | `context_max` | `int` | - | Maximum context window length in tokens | | `output_min` | `int` | - | Minimum max output length in tokens | | `output_max` | `int` | - | Maximum max output length in tokens | For array parameters, pass repeated query keys, for example `vendors=OpenAI&vendors=Google`. #### Example ```bash theme={null} curl "https://router-link.world3.ai/api/v1/models?page=1&page_size=20" ``` #### Response ```json theme={null} { "total": 66, "page": 1, "page_size": 1, "items": [ { "id": 58, "model_name": "Gemini 3.1 Pro Preview", "vendor": "Google", "providers": "WORLD3 Router North America", "description": "Gemini 3.1 Pro Preview is Google's frontier reasoning model...", "label": "Recommended", "tags": "Text, Image, File, Audio, Video", "total_context": 1050000, "max_output": 65500, "is_free": false, "created_at": "2026-03-31T15:18:37", "updated_at": "2026-03-31T15:18:37" } ] } ``` The values above are an example response shape. The live catalog can change. *** ### Model Group Returns the provider-channel implementations for a given logical model, grouped by RouterLink format standard and provider-channel alias. This powers model detail pages where users can see which channels can serve a model. ```http theme={null} GET /api/v1/models/group?model_name=Gemini%203.1%20Pro%20Preview ``` #### Response Shape ```json theme={null} { "Gemini 3.1 Pro Preview": { "OpenAI": { "WORLD3 Router North America": [ { "model_id": 1097, "model_name": "world3-router-north-america/google/gemini-3.1-pro-preview", "display_name": "Gemini 3.1 Pro Preview", "model_standard": "OpenAI", "provider": "WORLD3 Router North America", "endpoint": "https://router-link.world3.ai/api", "vendor": "Google", "total_context": 1050000, "max_output": 65500, "pricing": { "input": [ { "token_min": 0, "token_max": 200000, "price_usd": "2.20000000000000", "price_wai_credit": 2200 } ], "output": [ { "token_min": 0, "token_max": 200000, "price_usd": "13.20000000000000", "price_wai_credit": 13200 } ] } } ] } } } ``` Currently, the public `GET /api/v1/{model_standard}/models` route accepts RouterLink format standards such as `OpenAI` and `Anthropic`. Use `/api/v1/models/group` when you need the grouped view for a logical model. *** ### Providers List Returns all available provider-channel aliases, sorted alphabetically. Provider-channel aliases are routing channels, not necessarily model vendors. ```http theme={null} GET /api/v1/models/providers ``` ```bash theme={null} curl "https://router-link.world3.ai/api/v1/models/providers" ``` Example values include `OpenAI Channel`, `Anthropic Channel`, `GCP Vertex`, `WORLD3 Router North America`, and `WORLD3 Router Asia Pacific`. *** ### Vendors List Returns all unique vendor names, sorted alphabetically. Vendors are the model owners or model families, such as `OpenAI`, `Anthropic`, `Google`, `DeepSeek`, `Qwen`, and others. ```http theme={null} GET /api/v1/models/vendors ``` ```bash theme={null} curl "https://router-link.world3.ai/api/v1/models/vendors" ``` *** ## Model Labels Models on **routerlink.ai/models** are tagged with labels to help you choose the right one: | Label | Meaning | | -------------- | -------------------------------------------------------- | | `Recommended` | Best overall value for most use cases | | `Flagship` | Top-tier capability models from a vendor or model family | | `Free for Now` | Currently free to test (pricing subject to change) | | `Deprecated` | End-of-life models retained for backward compatibility | *** ## Pricing RouterLink stores pricing as per-model, per-provider-channel, per-token-type tiers in the `model_pricing` table. Pricing is returned by endpoints that expose model implementations, such as `GET /api/v1/models/group` and `GET /api/v1/{model_standard}/models?with_price=true`. Each pricing entry contains: | Field | Description | | ------------------ | ---------------------------------------------------------------- | | `token_min` | Inclusive lower bound for the usage tier | | `token_max` | Inclusive upper bound for the usage tier; `null` means unlimited | | `price_usd` | USD price per 1,000,000 tokens for this tier | | `price_wai_credit` | RouterLink credit price per 1,000,000 tokens for this tier | Supported token types are defined by the registry and currently include: | Token Type | Description | | --------------------------------------------------------------- | ---------------------------------- | | `input` | Input text tokens | | `output` | Output text tokens | | `cache_read` | Cached input tokens read | | `cache_write` | Cached input tokens written | | `input_audio` / `output_audio` | Audio input or output tokens | | `input_image` / `output_image` | Image input or output units/tokens | | `input_video` / `output_video` | Video input or output units/tokens | | `input_audio_cache` / `input_image_cache` / `input_video_cache` | Cached modality-specific input | | `output_per_audio` / `output_per_image` / `output_per_video` | Per-output media unit pricing | A price of `0` means the tier is free. *** ## Chat Completion Request Parameters When calling `POST /api/v1/chat/completions`, supported parameters depend on the target model and provider channel. Common OpenAI-compatible parameters include: | Parameter | Description | | -------------------------------------- | --------------------------------------------------------------------------------- | | `model` | Target RouterLink model identifier | | `messages` | Conversation array | | `temperature` | Output randomness | | `top_p` | Nucleus sampling | | `n` | Number of completions to generate | | `max_tokens` / `max_completion_tokens` | Maximum completion length | | `stop` | Custom stop sequences | | `stream` / `stream_options` | Streaming controls | | `presence_penalty` | Encourage topic diversity | | `frequency_penalty` | Reduce repetition | | `logit_bias` | Adjust token likelihoods | | `logprobs` / `top_logprobs` | Token log probability controls | | `tools` | Function/tool definitions | | `tool_choice` | Control tool selection behavior | | `parallel_tool_calls` | Allow or disable parallel tool calls | | `response_format` | Response format control, including JSON/schema-style outputs on supporting models | | `modalities` / `audio` | Multimodal output controls on supporting models | | `reasoning_effort` | Reasoning effort on supporting models | | `web_search_options` | Web search options on supporting models | | `metadata` | Caller-provided metadata | | `service_tier` | Provider-specific service tier | | `store` | Provider-side storage preference where supported | | `verbosity` | Verbosity control where supported | Provider-specific extra fields may be accepted by some channels, but portable behavior should be based on the parameters above and the target model's capabilities. *** ## Data Model RouterLink maintains a unified MySQL registry across three core tables: | Table | Purpose | | ---------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- | | `model` | Provider-channel implementation rows: model identifiers, provider channel, format standard, endpoint, context window, output limit, and active state | | `model_pricing` | Per-provider-channel, per-token-type pricing tiers | | `model_metadata` | Logical model metadata: vendor, provider list, tags, labels, descriptions, and display names | A fourth table, **`model_fallback_chains`**, defines provider failover behavior. When a primary provider is exhausted, the gateway can try the next enabled fallback entry ordered by `priority`. *** ## Providers and Vendors RouterLink separates provider channels from vendors: | Concept | Meaning | Example | | ---------------- | ---------------------------------------------------------- | ------------------------------------------------------------- | | Provider channel | The routing channel that serves a request | `WORLD3 Router North America`, `GCP Vertex`, `OpenAI Channel` | | Vendor | The model owner or model family | `OpenAI`, `Anthropic`, `Google`, `DeepSeek`, `Qwen` | | Format standard | The API compatibility shape used by a model implementation | `OpenAI`, `Anthropic` | Use `GET /api/v1/models/providers` and `GET /api/v1/models/vendors` to retrieve the current live lists.