Parameters & Models

Complete reference for request fields and model selection.

Common parameters

Parameter	Required	Type	Default	Description
model	Yes	string	—	Model id or model name
provider	No	string	model’s provider	Must match resolved provider if set
type	No	string	text	text, image, or embedding

The API resolves model by name first, then by id. Prefer the exact names returned by GET https://ai.lumowp.com/api/external/ai.

Text parameters

Parameter	Type	Default	Description
prompt	string	—	Required unless messages is provided
messages	array	—	Chat turns: { role, content }
system	string	—	Optional system prompt with prompt
max_tokens	number	capped	Maximum reply/output tokens to generate; it does not change the model’s context capacity
temperature	number	provider default	Creativity / randomness (0.0–2.0)
stream	boolean	false	SSE streaming when supported

Image parameters

Parameter	Type	Default	Description
prompt	string	—	Required image description
size	string	1024x1024	1024x1024, 1792x1024, 1024x1792 (avoid 256x256/512x512 for openai/gpt-image-1-mini)
quality	string	medium	low, medium, high (legacy: standard→medium, hd→high)
style	string	—	vivid or natural when supported
stream	boolean	false	Not supported for images

Embedding parameters

Parameter	Type	Default	Description
input	string or string[]	—	Required text(s) to embed
stream	boolean	false	Not supported for embeddings

Supported model families

Available models depend on your plan/credits, API key restrictions, and live provider configuration. Always discover the current list with:

curl -X GET "https://ai.lumowp.com/api/external/ai" \
  -H "X-API-Key: $LUMO_AI_API_KEY"

Typical families include:

Text / chat

Provider	Example models	Best for
Lumo AI	lumo-ai	Stable default alias; upstream routing is managed for you
OpenAI	gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, gpt-5.4, gpt-5.4-mini, gpt-5.4-nano	Complex reasoning, balanced, high-volume
Anthropic	claude-fable-5, claude-opus-4-8, claude-sonnet-5, claude-haiku-4-5	Agents, coding, fast tasks
Google	gemini-3.6-flash, gemini-3.5-flash-lite, gemini-2.5-pro	Multimodal, agentic, cost-efficient
Mistral	mistral-medium-3-5, mistral-large-latest, mistral-small-latest, codestral-latest	Multilingual, agentic, coding
Grok (xAI)	grok-4.5, grok-4.3, grok-build-0.1	Tool calling, coding

Image

Provider	Example models	Notes
OpenAI	openai/gpt-image-1-mini	Size / quality (low/medium/high) options; returns hosted URL
Google	gemini-3.1-flash-image, gemini-3.1-flash-lite-image, gemini-3-pro-image	Nano Banana family
Other image providers	See image_generation_providers	Capability flags in GET response

Embeddings

Provider	Example models	Notes
OpenAI	text-embedding-3-small, text-embedding-3-large	See embedding_providers + dimensions

Example names above are illustrative. If a name is not returned by GET on your site, it is unavailable.

gpt-5.4-pro is intentionally not listed: it requires OpenAI’s Responses API, while this service currently routes OpenAI text requests through Chat Completions.

Capability flags

Each model in the GET response includes:

{
  "id": "...",
  "name": "...",
  "capabilities": {
    "chat": true,
    "imageGeneration": false,
    "imageAnalysis": false,
    "streaming": true,
    "embedding": false
  },
  "context_length": 128000,
  "max_output_tokens": 8192,
  "maxTokens": 8192
}

context_length is the combined capacity for input, conversation history, and output. max_output_tokens is the largest reply this gateway will accept for that model. maxTokens is a deprecated compatibility alias of max_output_tokens.

Use these flags to filter UI options instead of hard-coding model lists.

Token costs

Token deduction is calculated from the selected model and request type:

• Text requests use input/output token pricing
• Image requests use the model’s image cost configuration
• Embedding requests use embedding pricing when configured

If pricing configuration is unavailable, the API returns 503 rather than guessing a cost.

Choosing a model

1. Call GET https://ai.lumowp.com/api/external/ai
2. Filter by capabilities.chat, capabilities.imageGeneration, or capabilities.embedding
3. Respect api_key_info.allowed_providers and allowed_models
4. Pass the chosen model (and matching provider) on POST