Media tools

LLM Router

Text and vision answers through the LLM Router — Gemini 3.8 Flash on a prompt, up to 10 images and one video, billed per token at list price.

LLM Router is the Sume model surface for language models. It is in the same family as Image Router and Video Router: one endpoint, a model id from the catalog, your Sume API key. Today the catalog has one model, gemini-3.8-flash.

Primary invoke URL:

text

POST /v1/llm-router/generate

Catalog:

text

GET /v1/llm-router/models
GET /v1/llm-router/models/{model_id}

The call answers when the model does (no job to poll).

Request fields

Field	Required	Notes
`model`	Yes	`gemini-3.8-flash`
`prompt`	Yes	The user message, up to 200,000 characters.
`system_prompt`	No	Instructions sent before the prompt.
`image_urls`	No	Up to 10 public HTTPS images (PNG, JPEG, WebP, HEIC), in order.
`video_url`	No	One public HTTPS video, up to 100 MB.
`temperature`	No	0–2.
`max_tokens`	No	Output cap, thinking included. Default 4,096, at most 32,768.
`response_format`	No	`text` (default) or `json`.

bash

curl -X POST https://api.sume.com/v1/llm-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3.8-flash",
    "system_prompt": "Answer in one sentence.",
    "prompt": "What happens in this clip?",
    "video_url": "https://media.sume.com/example.mp4"
  }'

Response

json

{
  "data": {
    "model": "gemini-3.8-flash",
    "output": "A dancer spins twice and bows.",
    "finish_reason": "STOP",
    "usage": {
      "prompt_tokens": 4210,
      "completion_tokens": 182,
      "reasoning_tokens": 160,
      "total_tokens": 4392,
      "cost": "$0.003840"
    }
  }
}

Price

Model	Input	Output
`gemini-3.8-flash`	$0.75 per 1M tokens	$3.75 per 1M tokens

Thinking tokens count as output. Images and video count as input tokens. The request holds 200,000 input tokens plus max_tokens of output from your balance while it runs, then charges the tokens the model reports and returns the rest. A failed call charges nothing.

Errors

Status	`code`	When
400	`invalid_request`	A field is wrong, or an image or the video cannot be read.
402	`insufficient_credits`	The balance does not cover the hold.
404	`model_not_found`	`GET /v1/llm-router/models/{model_id}` with an unknown id.
502	`provider_submission_failed`	The model failed or did not answer. Nothing is charged.
503	`provider_not_configured`	The model is not available on this host.

LLM Router

Text and vision answers through the LLM Router — Gemini 3.8 Flash on a prompt, up to 10 images and one video, billed per token at list price.

Primary invoke URL:

text

POST /v1/llm-router/generate

Catalog:

text

GET /v1/llm-router/models
GET /v1/llm-router/models/{model_id}

The call answers when the model does (no job to poll).

Request fields

Field	Required	Notes
`model`	Yes	`gemini-3.8-flash`
`prompt`	Yes	The user message, up to 200,000 characters.
`system_prompt`	No	Instructions sent before the prompt.
`image_urls`	No	Up to 10 public HTTPS images (PNG, JPEG, WebP, HEIC), in order.
`video_url`	No	One public HTTPS video, up to 100 MB.
`temperature`	No	0–2.
`max_tokens`	No	Output cap, thinking included. Default 4,096, at most 32,768.
`response_format`	No	`text` (default) or `json`.

bash

curl -X POST https://api.sume.com/v1/llm-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3.8-flash",
    "system_prompt": "Answer in one sentence.",
    "prompt": "What happens in this clip?",
    "video_url": "https://media.sume.com/example.mp4"
  }'

Response

json

{
  "data": {
    "model": "gemini-3.8-flash",
    "output": "A dancer spins twice and bows.",
    "finish_reason": "STOP",
    "usage": {
      "prompt_tokens": 4210,
      "completion_tokens": 182,
      "reasoning_tokens": 160,
      "total_tokens": 4392,
      "cost": "$0.003840"
    }
  }
}

Price

Model	Input	Output
`gemini-3.8-flash`	$0.75 per 1M tokens	$3.75 per 1M tokens

Errors

Status	`code`	When
400	`invalid_request`	A field is wrong, or an image or the video cannot be read.
402	`insufficient_credits`	The balance does not cover the hold.
404	`model_not_found`	`GET /v1/llm-router/models/{model_id}` with an unknown id.
502	`provider_submission_failed`	The model failed or did not answer. Nothing is charged.
503	`provider_not_configured`	The model is not available on this host.