AI Gateway
Models & pricing
The Clipia AI Gateway LLM catalog in GET /v1/models and prices in US dollars per 1M input and output tokens — Claude, GPT, Gemini, DeepSeek and Grok.
Clipia AI Gateway provides access to flagship language models under one contract. GET /v1/models returns the catalog in an OpenAI-compatible format; the prices below are in US dollars per 1M tokens, billed separately for input (prompt) and output (the model's response).
Catalog (GET /v1/models)
/v1/modelsmodels = client.models.list()
for m in models.data:
print(m.id)The response is a data array in OpenAI-compatible form. Each item carries an id (the string for the model field), context length, supported parameters and per-token rates:
{
"object": "list",
"data": [
{
"id": "gpt-6.1-sol",
"canonical_slug": "openai/gpt-6.1-sol",
"object": "model",
"name": "GPT-6.1 Sol",
"created": 0,
"owned_by": "openai",
"context_length": 272000,
"architecture": {
"modality": "text+image->text",
"input_modalities": ["text", "image"],
"output_modalities": ["text"],
"tokenizer": "GPT"
},
"pricing": { "prompt": "0.00000321", "completion": "0.00001605" },
"top_provider": { "context_length": 272000, "max_completion_tokens": 8192, "is_moderated": false },
"supported_parameters": [
"max_tokens", "tools", "tool_choice", "response_format",
"structured_outputs", "reasoning", "reasoning_effort"
],
"per_request_limits": null
}
]
}In pricing, values are in US dollars per single token (prompt / response), converted from the credit rate; the human-readable per-1M-token prices are in the table below. Clients use supported_parameters to detect which features a model supports (tools, response_format, etc.); created is 0 when the developer's release date is not confirmed, and top_provider.max_completion_tokens is the response length when max_tokens is not sent. The same list continues after the text models with image, video and audio models (owned_by: "clipia", with slug, type and a price in credits): text models differ by the developer in owned_by and by pricing in dollars per token. GET /v1/models/{id} returns one model; an unknown id returns 404.
Pricing
Billing is per token, separately for input and output. Charges are taken in credits from the account balance; the prices below are shown in US dollars for clarity (1M = 1,000,000 tokens).
| Model | model | Input, $/1M | Output, $/1M |
|---|---|---|---|
| Claude Fable 5.1 | claude-fable-5-1 | 16.05 | 80.24 |
| GPT-6 Astra | gpt-6-astra | 16.05 | 80.24 |
| GPT-6.1 Sol | gpt-6.1-sol | 3.21 | 16.05 |
| Claude Opus 5.5 | claude-opus-5-5 | 6.42 | 32.10 |
| Claude Sonnet 5.5 | claude-sonnet-5-5 | 3.21 | 16.05 |
| Gemini 3.8 Flash | gemini-3.8-flash | 1.50 | 7.50 |
| Grok 4.7 | grok-4.7 | 3.21 | 9.63 |
| DeepSeek V4.1 Flash | deepseek-v4.1-flash | 0.48 | 1.93 |
| GPT-5.6 Terra | gpt-5.6-terra | 3.21 | 19.26 |
| GPT-6 Luna | gpt-6-luna | 0.16 | 0.80 |
| DeepSeek V4 Pro | deepseek-v4-pro | 2.12 | 6.36 |
| Gemini 3.5 Flash Lite | gemini-3.5-flash-lite | 0.48 | 4.01 |
| GPT-5 nano | gpt-5.4-nano | 0.08 | 0.65 |
DeepSeek V4.1 Flash supports text chat and reasoning without tool calling. Grok 4.7 is limited to a 200,000-token context in Clipia. Check each model’s supported_parameters in GET /v1/models for its capabilities.
Previous model identifiers remain available for API compatibility and request history, so the GET /v1/models catalog is longer than this table. Use the current generations in the table for new integrations. Google retired gemini-2.5-flash on October 20, 2026, so requests with this identifier are served by gemini-3.5-flash-lite: the response model field names the model that answered, and you are charged at its price. In new integrations, set a current identifier explicitly.
Pricing
Prices are current as of November 2026. The API returns the cost of each request in credits by the response id through GET /v1/generation — see Key and limits; the response's own usage holds token counts only.
The model list and rates change over time. Take identifiers and current prices from GET /v1/models — that's the source of truth; the table above is for orientation.
Context and response length
The context is the input cap: the server rejects a larger request with 400 context_length_exceeded before anything is charged. The response length is the value of max_tokens when you do not send it; do not send more than that. In GET /v1/models, the same numbers are in context_length and top_provider.max_completion_tokens.
| Model | Context, tokens | Default response, tokens |
|---|---|---|
| Claude Fable 5.1 | 1,000,000 | 8,192 |
| GPT-6 Astra | 1,050,000 | 8,192 |
| GPT-6.1 Sol | 272,000 | 8,192 |
| Claude Opus 5.5 | 1,000,000 | 8,192 |
| Claude Sonnet 5.5 | 1,000,000 | 8,192 |
| Gemini 3.8 Flash | 1,048,576 | 8,192 |
| Grok 4.7 | 200,000 | 8,192 |
| DeepSeek V4.1 Flash | 1,000,000 | 8,192 |
| GPT-5.6 Terra | 400,000 | 8,192 |
| GPT-6 Luna | 272,000 | 8,192 |
| DeepSeek V4 Pro | 128,000 | 4,096 |
| Gemini 3.5 Flash Lite | 1,048,576 | 8,192 |
| GPT-5 nano | 128,000 | 2,048 |
How to choose a model
- Flagships — hard tasks.
claude-fable-5-1,gpt-6-astra,gpt-6.1-solandclaude-opus-5-5for reasoning, code, agents and long context, where quality matters more than price. - Quality/price balance.
claude-sonnet-5-5,gemini-3.8-flashandgrok-4.7are all-rounders for most product scenarios. - Volume and low cost.
gpt-6-luna,deepseek-v4.1-flash,gemini-3.5-flash-liteandgpt-5.4-nanofor classification, data extraction, bulk processing and draft answers.
Start with a mid-tier model, measure quality on your own data, and move up to a flagship or down to a cheaper option as needed — you only change the model string.
Previous-generation deep dives
Reference pages for previous generations. Choose current identifiers from the table above for new requests:
- Claude Opus 5 — 1,000,000-token context, for deep analysis, code and agent workloads.
- Gemini 3.6 Flash — 1,048,576-token context, multimodal input and a low per-request cost.
Chat Completions
POST /v1/chat/completions — the OpenAI-compatible chat endpoint of Clipia AI Gateway. Streaming and regular responses, parameters, roles, tools and errors.
LangChain
Connect Clipia AI Gateway to LangChain via ChatOpenAI with the Clipia base_url — chat, tool calling, agents and structured output on Claude, GPT and Gemini.