List Models

GET /v1/models Returns the list of models currently available through Consus Gateway.

Request

Headers

Header	Required	Description
`x-api-key`	Yes	Your API key

Example

curl https://api.consus.io/v1/models \
  -H "x-api-key: $CONSUS_API_KEY"

Response

The response lists every requestable composite ID across all active models. Each base model expands into its full ladder of authorized levels (FedRAMP Low/Moderate/High plus DoD IL2/IL4/IL5 where applicable, with +itar variants when the underlying provider is ITAR authorized). ITAR-authorized models additionally list a standalone :itar ID (ITAR with no compliance level). IDs are listed with the consus/ prefix; requests accept an ID with or without it. Each entry also carries the model’s limits and price, so a client can configure itself from this list: the context window, the largest max_tokens the model accepts, which inputs it takes, which reasoning efforts it honors, and what each token costs at that ID’s level. The example below is truncated to four entries.

{
  "object": "list",
  "data": [
    {
      "id": "consus/gpt-5.6-sol:il5+itar",
      "object": "model",
      "created": 1700000000,
      "owned_by": "openai",
      "display_name": "GPT-5.6 Sol",
      "context_window": 922000,
      "max_output_tokens": 128000,
      "input_modalities": ["text", "image"],
      "reasoning_efforts": ["low", "medium", "high", "xhigh"],
      "pricing": { "input": 5.5, "output": 27.5, "cache_read": 0.55, "cache_write": 0.0 }
    },
    {
      "id": "consus/claude-sonnet-4-5:il5",
      "object": "model",
      "created": 1700000000,
      "owned_by": "anthropic",
      "display_name": "Claude Sonnet 4.5",
      "context_window": 200000,
      "max_output_tokens": 64000,
      "input_modalities": ["text", "image", "pdf"],
      "reasoning_efforts": ["low", "medium", "high"],
      "pricing": { "input": 3.6, "output": 18.0, "cache_read": 0.36, "cache_write": 4.5 }
    },
    {
      "id": "consus/grok-4.6:il5",
      "object": "model",
      "created": 1700000000,
      "owned_by": "xai",
      "display_name": "Grok 4.6",
      "context_window": 500000,
      "max_output_tokens": 500000,
      "input_modalities": ["text", "image"],
      "reasoning_efforts": ["low", "medium", "high", "xhigh"],
      "pricing": { "input": 2.64, "output": 7.92, "cache_read": 0.66, "cache_write": 0.0 }
    },
    {
      "id": "consus/gpt-oss-120b:il5",
      "object": "model",
      "created": 1700000000,
      "owned_by": "openai",
      "display_name": "GPT OSS 120B",
      "context_window": 128000,
      "max_output_tokens": 16384,
      "input_modalities": ["text"],
      "reasoning_efforts": ["low", "medium", "high"],
      "pricing": { "input": 0.18, "output": 0.72, "cache_read": 0.0, "cache_write": 0.0 }
    }
  ]
}

GPT-5.6 Sol is served from Azure Government. Sol is the one GPT-5.6 model on Azure Government OpenAI (FedRAMP High, DoD IL5, and ITAR — the same contract as GPT-5.1), exposing the full id ladder: :fedramp-high, :il5, lower levels, level+itar composites, and the standalone :itar id. It is available on /v1/chat/completions and /v1/responses with 922K input context, text + image input, tool calling, and reasoning_effort up to xhigh (the gateway serves Sol’s chat-completions traffic through Azure’s Responses API upstream, as it does for GPT-5.6 on Bedrock, so tools and reasoning work together). GPT-5.6 Luna and Terra remain on AWS GovCloud Bedrock. Pricing is $5.50 / $27.50 per 1M tokens in/out, cache reads at $0.55, matching Azure Government’s current Data Zone rate. Requests above 272K input tokens are billed at a long-context rate — see the GPT-5.6 pricing note below.

Claude Opus 5 is served from two clouds. FedRAMP-level and :il2 requests (claude-opus-5:fedramp-high, :il2, and the lower FedRAMP ranks) route to Vertex AI in GCP Assured Workloads; ITAR requests (:itar and the fedramp-*+itar composites) route to AWS Bedrock GovCloud. Neither deployment is authorized above DoD IL2, so claude-opus-5:il4 and :il5 return 400 not available. Token counting (/v1/messages/count_tokens) works on the Vertex-served ids and fails closed on the Bedrock-served ones — see Messages.

Claude Opus 5.5 is served from two clouds, routed the same way as Opus 5. Opus 5.5 offers a 1M-token context window, 128K max output, and text + image input, and is served on both /v1/chat/completions (where reasoning_effort from low to xhigh sets the thinking depth) and /v1/messages (which takes the native thinking block instead). Extended thinking is always on and cannot be disabled — see Chat Completions. Pricing per 1M tokens in/out is $4.40 / $22.00 on Vertex (cache reads $0.22) and $4.80 / $24.00 on Bedrock (cache reads $0.24).

Claude Sonnet 5.5 is served from AWS Bedrock GovCloud at FedRAMP High and ITAR. The ids are claude-sonnet-5-5:fedramp-high (and the lower FedRAMP ranks), the fedramp-*+itar composites, and the standalone claude-sonnet-5-5:itar. AWS has not authorized it at a DoD impact level, so claude-sonnet-5-5:il2, :il4, and :il5 return 400 not available — use Claude Sonnet 5 when you need IL5. Sonnet 5.5 offers a 1M-token context window, 128K max output, and text + image input, and is served on both /v1/chat/completions (where reasoning_effort from low to xhigh sets the thinking depth; the default is high) and /v1/messages (which takes the native thinking block instead). Extended thinking is always on and cannot be disabled — see Chat Completions. Token counting is not yet available on any of its ids — see Messages. Pricing per 1M tokens in/out is $2.40 / $12.00, cache reads $0.12.

Claude Haiku 5.5 is served from two clouds. The FedRAMP ids (claude-haiku-5-5:fedramp-high and the lower ranks) route to GCP Vertex AI at FedRAMP High; claude-haiku-5-5:itar routes to AWS Bedrock GovCloud. There is no DoD impact level and no fedramp-*+itar composite, so those ids return 400 — use :itar for export-controlled work. 1M-token context, 128K max output, text + image input, on /v1/chat/completions (reasoning_effort low–xhigh, default high) and /v1/messages, where "thinking": {"type": "disabled"} turns thinking off at effort low–high. Token counting works on the FedRAMP ids only — see Messages. Pricing per 1M tokens in/out is tiered by prompt length on both clouds: Vertex $0.11 / $0.55 (cache reads $0.011) and Bedrock $0.12 / $0.60 (cache reads $0.012) up to 100,000 input tokens, and 5× for the whole request above that (Vertex $0.55 / $2.75, cache reads $0.055; Bedrock $0.60 / $3.00, cache reads $0.06).

Grok 4.6 (xAI) is authorized to FedRAMP High / DoD IL5. It is served from AWS Bedrock GovCloud and exposes the full ladder of ids — grok-4.6:fedramp-high, grok-4.6:il5, lower levels, level+itar composites, and the standalone grok-4.6:itar id. Grok 4.6 is available on /v1/chat/completions only (500K context, text + image input, reasoning_effort from low to xhigh); /v1/messages and /v1/responses reject it.

Gemini 4 Argon is authorized to FedRAMP High / DoD IL4. Like every Gemini model it is served on Google’s US multi-region endpoint: :fedramp-* ids from the FedRAMP High project, :il4 / :il2 from the IL5 project. It is not ITAR-authorized, so :il5 and +itar ids return 400. 1M-token context and output, /v1/chat/completions only, reasoning_effort none–high (default medium). Introductory pricing through 2026-11-30 is $2.00 / $10.00 per 1M in/out ($0.10 cached), rising to $4.00 / $20.00 / $0.20 on 2026-12-01.

Gemini 3 Flash and Gemini 3.1 Pro were retired on 2026-10-13. Both previews exist only on Google’s global endpoint, which carries no US data-residency commitment; their ids now return 400. Use Gemini 3.8 Flash or Gemini 4 Argon.

GPT-5.6 Luna and Terra are authorized to FedRAMP High / DoD IL5. Both models expose the full ladder of ids — :fedramp-high, :il5, lower levels, level+itar composites, and the standalone :itar id — the same contract as GPT-5.4 on the same Bedrock plane.

GPT-5.6 pricing: short and long context (Luna, Terra, Sol). AWS GovCloud and Azure Government price the GPT-5.6 family in two tiers based on each request’s input size. Starting 2026-10-01 00:00 UTC, Consus bills the same way: a request whose total input exceeds 272,000 tokens (cached tokens included) is billed entirely at the long-context rate — input, cached input, and output alike. Output length never changes the tier. Requests at or under 272,000 input tokens are unaffected. Before that date every request bills at the short-context rate.

Model	Short context (≤272K input) in / out / cached	Long context (>272K input) in / out / cached
`gpt-5.6-luna`	$0.27 / $1.62 / $0.027	$0.54 / $2.43 / $0.054
`gpt-5.6-terra`	$2.70 / $16.20 / $0.27	$5.40 / $24.30 / $0.54
`gpt-5.6-sol`	$5.50 / $27.50 / $0.55	$11.00 / $41.25 / $1.10

Prices are per 1M tokens. The cost of each request in your usage logs reflects the tier it was billed at. The context window itself is unchanged: 1M tokens for Luna and Terra, 922K for Sol.

Model Object

Field	Type	Description
`id`	string	The model identifier to use in API requests
`object`	string	Always `"model"`
`created`	integer	Unix timestamp
`owned_by`	string	The organization that created the model (e.g., `anthropic`, `google`)
`display_name`	string	Human-readable model name. It does not include the compliance level; read that from `id`.
`context_window`	integer	Maximum input tokens
`max_output_tokens`	integer or null	The largest `max_tokens` the model accepts. `null` for embedding models.
`input_modalities`	string[]	Accepted inputs: `text`, `image`, `pdf`, `video`. Enforced on `/v1/chat/completions`: an `image_url` part needs `image`, a `file` part needs `pdf`; a mismatch returns `400` with `code: "unsupported_input_modality"` before any provider call.
`reasoning_efforts`	string[]	The `reasoning_effort` values the model honors, weakest first (from `low`, `medium`, `high`, `xhigh`). `[]` means the model doesn’t reason; a `reasoning_effort` sent to it is accepted and ignored. A value above the last entry is lowered to it rather than rejected.
`pricing`	object	USD per 1M tokens for this ID; see below

pricing

Field	Type	Description
`input`	number	Input tokens
`output`	number	Output tokens
`cache_read`	number	Cache reads; `0` when the model has no cache-read rate
`cache_write`	number	Cache writes (5-minute); `0` when the model has no cache-write rate

Pricing is per ID because some models are served from more than one cloud and the compliance level decides which one handles the request; for example, claude-sonnet-4-5:fedramp-high and claude-sonnet-4-5:il5 list different rates. The standard rate is shown; GPT-5.6 long-context rates are in the note above. Only models with active provider integrations are returned. As new providers are integrated, their models will appear automatically.

Product

Getting Started

API Reference

Integrations

Request

Headers

Example

Response

Model Object

​Request

​Headers

​Example

​Response

​Model Object

Request

Headers

Example

Response

Model Object