Pricing

Prices are on the model catalog and GET /v1/models/{model_id}.

usage.cost: that call’s charge in USD, to six decimals. 0.0 is a charge of zero.
Included in: chat, including streamed chat, embeddings, and transcription.
Transcription: uses input_audio_seconds.

Billing variations

USD per 1 million tokens, unless that model’s catalog page says otherwise. method and response_format do not change the price.

By tokens
By images
By seconds

Most models bill per token. For example, google/gemma-4-26b-a4b-it bills $0.10 / 1M input tokens and $0.30 / 1M output tokens.

Field	Bills
`prompt`	Uncached input tokens
`input_cache_read`	Cached input tokens
`input_cache_write`	Always $0.00
`completion`	Output tokens

facebook/sam3.1 bills per image (for example, $1 / 1K images).

Field	Bills
`image`	One image, or one PDF page
`prompt`	$0.00 for facebook/sam3.1
`video`	Once per video, when the rate is set

Transcription models bill by audio length. For example, nvidia/parakeet-tdt-0.6b-v3 bills $16.67 / 1M audio seconds (about $0.06 per hour of audio).

Field	Bills
`prompt`	USD per 1 million audio seconds. Listed on `GET /v1/openai/models`
`completion`	$0.00

json and verbose_json are billed. text, srt, and vtt return the transcript only.

Billing transparency

Rates are published, not quoted. You can read what a model costs before you call it, then check what a call cost after.

Before the call: every model lists its rates in pricing on GET /v1/openai/models and GET /v1/models/{model_id}. The same rates appear on the model catalog.
After the call: usage.cost is the charge for that request in USD.

The unit in pricing follows how the model bills: tokens, images, or audio seconds.

By tokens
By images
By seconds

For the example below, google/gemma-4-26b-a4b-it is priced in USD per 1 million tokens.

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.vlm.run/v1/openai",
    api_key="<VLMRUN_API_KEY>",
)
model = next(m for m in client.models.list() if m.id == "google/gemma-4-26b-a4b-it")
print(model.pricing)
# {'prompt': 0.1, 'completion': 0.3, 'input_cache_read': 0.05, 'input_cache_write': 0.0, 'image': 0.0, 'video': 0.0}

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://gateway.vlm.run/v1/openai",
  apiKey: process.env.VLMRUN_API_KEY,
});
const models = await client.models.list();
const model = models.data.find((m) => m.id === "google/gemma-4-26b-a4b-it");
console.log(model.pricing);
// {
//   prompt: 0.1,
//   completion: 0.3,
//   input_cache_read: 0.05,
//   input_cache_write: 0,
//   image: 0,
//   video: 0,
// }

curl -s "https://gateway.vlm.run/v1/models/google%2Fgemma-4-26b-a4b-it" \
  -H "Authorization: Bearer $VLMRUN_API_KEY" | jq .pricing
# {
#   "prompt": 0.1,
#   "completion": 0.3,
#   "input_cache_read": 0.05,
#   "input_cache_write": 0.0,
#   "image": 0.0,
#   "video": 0.0
# }

For the example below, facebook/sam3.1 is priced in USD per 1 million images or clips. prompt and completion are 0.0.

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.vlm.run/v1/openai",
    api_key="<VLMRUN_API_KEY>",
)
model = next(m for m in client.models.list() if m.id == "facebook/sam3.1")
print(model.pricing)
# {'prompt': 0.0, 'completion': 0.0, 'input_cache_read': 0.0, 'input_cache_write': 0.0, 'image': 1000.0, 'video': 34800.0}

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://gateway.vlm.run/v1/openai",
  apiKey: process.env.VLMRUN_API_KEY,
});
const models = await client.models.list();
const model = models.data.find((m) => m.id === "facebook/sam3.1");
console.log(model.pricing);
// {
//   prompt: 0,
//   completion: 0,
//   input_cache_read: 0,
//   input_cache_write: 0,
//   image: 1000,
//   video: 34800,
// }

curl -s "https://gateway.vlm.run/v1/models/facebook%2Fsam3.1" \
  -H "Authorization: Bearer $VLMRUN_API_KEY" | jq .pricing
# {
#   "prompt": 0.0,
#   "completion": 0.0,
#   "input_cache_read": 0.0,
#   "input_cache_write": 0.0,
#   "image": 1000.0,
#   "video": 34800.0
# }

For the example below, nvidia/parakeet-tdt-0.6b-v3 is priced in USD per 1 million audio seconds, in prompt.

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.vlm.run/v1/openai",
    api_key="<VLMRUN_API_KEY>",
)
model = next(m for m in client.models.list() if m.id == "nvidia/parakeet-tdt-0.6b-v3")
print(model.pricing)
# {'prompt': 16.67, 'completion': 0.0, 'input_cache_read': 16.67, 'input_cache_write': 0.0, 'image': 0.0, 'video': 0.0}

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://gateway.vlm.run/v1/openai",
  apiKey: process.env.VLMRUN_API_KEY,
});
const models = await client.models.list();
const model = models.data.find((m) => m.id === "nvidia/parakeet-tdt-0.6b-v3");
console.log(model.pricing);
// {
//   prompt: 16.67,
//   completion: 0,
//   input_cache_read: 16.67,
//   input_cache_write: 0,
//   image: 0,
//   video: 0,
// }

curl -s "https://gateway.vlm.run/v1/models/nvidia%2Fparakeet-tdt-0.6b-v3" \
  -H "Authorization: Bearer $VLMRUN_API_KEY" | jq .pricing
# {
#   "prompt": 16.67,
#   "completion": 0.0,
#   "input_cache_read": 16.67,
#   "input_cache_write": 0.0,
#   "image": 0.0,
#   "video": 0.0
# }

Model catalog

Current rates for every model.

Get Model

Full response schema, including pricing.

Overview

Platform

OpenAI Compatibility

TypeSafe Compatibility

MCP

Misc

Billing variations

Billing transparency

Model catalog

Get Model

​Billing variations

​Billing transparency

Model catalog

Get Model

Billing variations

Billing transparency