--- title: "List all available models" method: GET path: "/public/v1/models" tags: ["Public Models"] --- # List all available models `GET /public/v1/models` List all publicly available models (modelVisibility='PUBLIC' only). Use the `format` parameter to get responses compatible with OpenRouter or HuggingFace APIs. ## Query parameters - `format` 'default' | 'openrouter' | 'huggingface' — Output format for public models endpoint. ## Headers - `X-Amz-Cf-Id` string, nullable - `CF-RAY` string, nullable ## Response `200` Successful Response - union - PublicModelsListResponse — OpenAI-compatible list of models. - `data` PublicModel[], required — Array of model objects - `architecture` ModelArchitecture, required — Architecture details of the model. - `instruct_type` string, nullable — The instruction format type used for fine-tuning (e.g., 'llama3', 'chatml'). - `modality` 'text' | 'text+vision' | 'multimodal', required — The modality of the model (e.g., 'text', 'text+vision', 'multimodal'). - `tokenizer` string, required — The tokenizer used by the model (e.g., 'Llama3', 'GPT4'). - `capabilities` ModelCapabilities, required — Capabilities and features supported by the model. - `function_calling` boolean — Indicates if the model supports function calling (tool use). - `json_mode` boolean — Indicates if the model supports JSON mode (guaranteed JSON output). - `parallel_tool_calls` boolean — Indicates if the model supports parallel tool calls. - `reasoning` boolean — Indicates if the model supports reasoning/chain-of-thought outputs. - `response_format` boolean — Indicates if the model supports the response_format parameter. - `streaming` boolean — Indicates if the model supports streaming responses via Server-Sent Events (SSE). - `structured_outputs` boolean — Indicates if the model supports structured outputs (e.g. JSON schema enforcement). - `tool_choice` boolean — Indicates if the model supports the tool_choice parameter. - `tools` boolean — Indicates if the model supports the tools parameter. - `vision` boolean — Indicates if the model accepts image inputs (vision capabilities). - `created` integer, required — The Unix timestamp (in seconds) when the model was created. - `datacenter_locations` string[] — List of datacenter locations where this model is deployed (e.g., ['us-east-1', 'eu-west-1']). - `deprecated` boolean — Indicates if the model is deprecated and should not be used for new applications. - `description` string, required — A brief description of the model. - `hugging_face_id` string, nullable — The corresponding HuggingFace Hub model ID, if available (e.g., 'meta-llama/Llama-3.1-8B-Instruct'). - `id` string, required — The unique identifier for the model (e.g., 'llama3.1-8b'). - `limits` ModelLimits, required — Rate limits and constraints for the model. - `max_completion_tokens` integer, required — The maximum number of tokens that can be generated in a single completion. - `max_context_length` integer, required — The maximum context window size in tokens. - `requests_per_minute` integer, nullable — The default rate limit for requests per minute (RPM). - `tokens_per_minute` integer, nullable — The default rate limit for tokens per minute (TPM). - `name` string, required — The human-readable name of the model. - `object` 'model' — The object type, which is always 'model'. - `owned_by` string, required — The organization that owns or created the model. - `preview` boolean — Indicates if the model is in preview or beta status. - `pricing` ModelPricing, required — Pricing information for a model. - `completion` string, required — Cost per token for completion (output) tokens in USD. - `prompt` string, required — Cost per token for prompt (input) tokens in USD. - `quantization` string, nullable — Quantization precision (e.g., 'FP16', 'FP16/FP8 (weights only)'). - `supported_parameters` SupportedParameters, required — Sampling parameters supported by the model. - `frequency_penalty` boolean — Supports frequency_penalty parameter. - `logit_bias` boolean — Supports logit_bias parameter. - `logprobs` boolean — Supports logprobs output. - `max_completion_tokens` boolean — Supports max_completion_tokens parameter. - `presence_penalty` boolean — Supports presence_penalty parameter. - `repetition_penalty` boolean — Supports repetition_penalty parameter. - `seed` boolean — Supports seed for reproducible outputs. - `stop` boolean — Supports stop sequences parameter. - `temperature` boolean — Supports temperature sampling parameter. - `top_logprobs` boolean — Supports top_logprobs parameter. - `top_p` boolean — Supports top_p (nucleus) sampling parameter. - `object` 'list' — Object type (always 'list') - OpenRouterModelsResponse — OpenRouter-compatible list of models. - `data` OpenRouterModel[], required - `context_length` integer, required - `created` integer, required — Unix timestamp when model was created - `datacenters` DatacenterLocation[] — Datacenter locations - `country_code` string, required — ISO 3166 Alpha-2 country code - `description` string — Model description - `hugging_face_id` string — The corresponding HuggingFace Hub model ID, if available - `id` string, required — Model ID with provider prefix, e.g., 'cerebras/llama3.1-8b' - `input_modalities` string[] — Supported input modalities (text, image, file) - `max_output_length` integer, required — Maximum number of output tokens - `name` string, required - `openrouter` OpenRouterInfo — OpenRouter metadata. - `slug` string, required — OpenRouter slug for the model - `output_modalities` string[] — Supported output modalities (text, image, file) - `pricing` OpenRouterPricing, required — OpenRouter pricing format. - `completion` string, required — Cost per output token as string - `image` string — Cost per image as string - `input_cache_read` string — Cost per cached input token read as string - `input_cache_write` string — Cost per cached input token write as string - `prompt` string, required — Cost per input token as string - `request` string — Cost per request as string - `quantization` string — Model quantization (fp16 only for Cerebras) - `supported_features` string[] — List of supported features - `supported_sampling_parameters` string[] — List of supported sampling parameters - HuggingFaceModelsResponse — HuggingFace-compatible list of models. - `data` HuggingFaceModel[], required - `capabilities` HuggingFaceCapabilities — HuggingFace capabilities format. - `function_calling` boolean - `streaming` boolean - `structured_outputs` boolean - `vision` boolean - `context_length` integer, required — Supported context length in tokens - `created` integer, required - `hugging_face_id` string, nullable — The corresponding HuggingFace Hub model ID - `id` string, required - `object` 'model' - `owned_by` string, required - `pricing` HuggingFacePricing, required — HuggingFace pricing format - price in USD per million tokens. - `input` number, required — Price in USD per million input tokens - `output` number, required — Price in USD per million output tokens - `object` 'list' ## Other responses - `422` — Validation Error ## Changes > 51 revisions in range; 49 not diffed. - **2026-07-08** `130749c46caf` — 6 breaking, 1 warning, 45 info - the `query` request parameter `format` was restricted to a list of enum values - the `anyOf[subschema #1: PublicModelsListResponse]/data/items/architecture` response's property type/format changed from ``/`` to `object`/`` for status `200` - the `anyOf[subschema #1: PublicModelsListResponse]/data/items/capabilities` response's property type/format changed from ``/`` to `object`/`` for status `200` - the `anyOf[subschema #1: PublicModelsListResponse]/data/items/limits` response's property type/format changed from ``/`` to `object`/`` for status `200` - …48 more [Change history](https://skmtc.dev/cerebras/apis/cerebras-cloud-api/changes/public/v1/models/get.md) --- [API](https://skmtc.dev/cerebras/apis/cerebras-cloud-api.md) · [All operations](https://skmtc.dev/cerebras/apis/cerebras-cloud-api/llms.txt) · [OpenAPI document](https://skmtc-service-production.skmtc.workers.dev/v1/apis/cerebras/cerebras-cloud-api/revisions/130749c46caf/schema)