Models

#What the response includes

Identity & access

• id — Model identifier used in API requests
• owned_by — Provider name (e.g. openai, anthropic, google)
• endpoint — Which API endpoint this model works with
• token_multiplier — How many tokens count against your daily limit per actual token
• premium — Whether this model requires a paid plan
• required_plan — Minimum plan tier needed (if premium)

Capability metadata (pulled from the model's upstream catalog so clients can configure limits dynamically)

• context_window — Maximum input context in tokens (integer | null)
• max_output_tokens — Maximum tokens the model can generate per response (integer | null)
• supports_vision, supports_tools / supports_function_calling, supports_reasoning, supports_json_mode, supports_audio_input, supports_image_output, supports_streaming — capability flags (boolean | null)
• input_modalities, output_modalities — array<string> | null (e.g. ["text", "image"])
• modality — raw upstream modality string (e.g. "text+image->text") or null
• tokenizer — tokenizer name when known, otherwise null
• description — short model description from the upstream catalog, or null
• pricing — { prompt, completion, image, request } per-token price strings (any field may be null), or null overall
• metadata_source — which upstream provided the metadata: "openrouter", "poe", or null
• metadata_status — "known" when capability metadata was resolved, "unknown" otherwise

> Unknown values are always returned as null (never 0). Treat null as "unknown" rather than as a real limit — for example, do not interpret "max_output_tokens": null as "this model can output 0 tokens".

#Code examples

1curl https://api.navy/v1/models

#Why it matters

• Populate model selectors dynamically
• Hide plan-locked models for the wrong tier
• Show endpoint compatibility before a request is sent