One unified API, Any Visual Model.

Document OCR, captioning, and multi-modal chat: every visual capability behind one MCP server.

Supported providers

  • Google
  • OpenAI
  • Anthropic
  • xAI
  • Meta
  • Moonshot
  • Microsoft
  • Qwen
  • DeepSeek
  • ZAI
  • Baidu
  • Roboflow
  • LightOn
  • RedNote
  • USYD
  • Hugging Face
Explore models

30 visual models. One endpoint.

Qwen3.8 27B

qwen/qwen3.8-27b

Usage
97K requests in the last 14 days
Context
164K
Input / 1M
$0.35
Cached / 1M
$0.05
Output / 1M
$2.55

SAM 3.1

facebook/sam3.1

Usage
8.84K requests in the last 14 days
Context
—
Input / 1M
$1/1K
Cached / 1M
—
Output / 1M
—

PP-OCRv6

paddleocr/pp-ocrv6

Usage
8.02K requests in the last 14 days
Context
—
Input / 1M
$0.02
Cached / 1M
$0.02
Output / 1M
$0.40

Qwen3.5 0.8B

qwen/qwen3.5-0.8b

Usage
7.14K requests in the last 14 days
Context
80K
Input / 1M
$0.08
Cached / 1M
$0.02
Output / 1M
$0.15

ViTPose+ Large

usyd-community/vitpose-plus-large

Usage
6.81K requests in the last 14 days
Context
—
Input / 1M
$1/1K
Cached / 1M
—
Output / 1M
—

DiffusionGemma 26B-A4B

google/diffusiongemma-26b-a4b-it

Usage
4.01K requests in the last 14 days
Context
80K
Input / 1M
$0.35
Cached / 1M
$0.05
Output / 1M
$1.50

Gemini 3.8 Flash

google/gemini-3.8-flash

Usage
1.38K requests in the last 14 days
Context
1M
Input / 1M
$0.75
Cached / 1M
$0.075
Output / 1M
$3.75

dots.mocr

rednote-hilab/dots.mocr

Usage
1.34K requests in the last 14 days
Context
24K
Input / 1M
$0.20
Cached / 1M
$0.03
Output / 1M
$0.40

Florence-2

microsoft/florence-2-base-ft

Usage
780 requests in the last 14 days
Context
—
Input / 1M
$0
Cached / 1M
—
Output / 1M
—

Gemini 4 Argon

google/gemini-4-argon

Usage
774 requests in the last 14 days
Context
1M
Input / 1M
$2
Cached / 1M
$0.10
Output / 1M
$10

Gemma 4 26B-A4B Instruct

google/gemma4-26b-a4b-it

Usage
762 requests in the last 14 days
Context
80K
Input / 1M
$0.10
Cached / 1M
$0.05
Output / 1M
$0.30

Gemini 3.5 Flash Lite

google/gemini-3.5-flash-lite

Usage
697 requests in the last 14 days
Context
1M
Input / 1M
$0.30
Cached / 1M
$0.03
Output / 1M
$2.50

PaddleOCR-VL 1.6

paddlepaddle/paddleocr-vl-1.6

Usage
694 requests in the last 14 days
Context
16K
Input / 1M
$0.15
Cached / 1M
$0.15
Output / 1M
$0.35

RF-DETR Large

roboflow/rfdetr-large

Usage
411 requests in the last 14 days
Context
—
Input / 1M
$1/1K
Cached / 1M
—
Output / 1M
—

Gemini 3.7 Flash

google/gemini-3.7-flash

Usage
398 requests in the last 14 days
Context
1M
Input / 1M
$0.75
Cached / 1M
$0.075
Output / 1M
$3.75

LightOnOCR-3 4B

lightonai/lightonocr-3-4b

Usage
374 requests in the last 14 days
Context
16K
Input / 1M
$0.20
Cached / 1M
$0.20
Output / 1M
$0.50

RF-DETR Seg Large

roboflow/rfdetr-seg-large

Usage
371 requests in the last 14 days
Context
—
Input / 1M
$3/1K
Cached / 1M
—
Output / 1M
—

Gemini 4 Carbon

google/gemini-4-carbon

Usage
342 requests in the last 14 days
Context
1M
Input / 1M
$2
Cached / 1M
$0.10
Output / 1M
$10

OCR Auto

vlm-run/ocr:auto

Usage
230 requests in the last 14 days
Context
—
Input / 1M
—
Cached / 1M
—
Output / 1M
—

GLM-OCR

zai-org/glm-ocr

Usage
156 requests in the last 14 days
Context
8K
Input / 1M
$0.10
Cached / 1M
$0.02
Output / 1M
$0.20

Kimi K3

moonshotai/kimi-k3

Usage
140 requests in the last 14 days
Context
256K
Input / 1M
$3
Cached / 1M
$0.30
Output / 1M
$15

HaMeR

geopavlakos/hamer

Usage
132 requests in the last 14 days
Context
—
Input / 1M
$0
Cached / 1M
—
Output / 1M
—

MiniMax M3

minimax/minimax-m3

Usage
127 requests in the last 14 days
Context
1M
Input / 1M
$0.30
Cached / 1M
$0.06
Output / 1M
$1.20

Gemini Robotics-ER 2 (preview)

google/gemini-robotics-er-2-preview

Usage
92 requests in the last 14 days
Context
128K
Input / 1M
$1
Cached / 1M
$0.10
Output / 1M
$5

Muse Spark 1.2

meta/muse-spark-1.2

Usage
80 requests in the last 14 days
Context
1M
Input / 1M
$1.25
Cached / 1M
$0.15
Output / 1M
$4.25

Muse Glimmer 30B

meta/muse-glimmer-30b

Usage
71 requests in the last 14 days
Context
128K
Input / 1M
$0.35
Cached / 1M
$0.04
Output / 1M
$1.50

EmbeddingGemma 2

google/embeddinggemma-2

Usage
36 requests in the last 14 days
Context
8K
Input / 1M
$0.018
Cached / 1M
$0.018
Output / 1M
—

Nemotron 3.5 Parse

nvidia/nemotron-3.5-parse

Usage
34 requests in the last 14 days
Context
16K
Input / 1M
$0.10
Cached / 1M
$0.02
Output / 1M
$0.20

Qwen3-VL Embedding 2B

qwen/qwen3-vl-embedding-2b

Usage
22 requests in the last 14 days
Context
8K
Input / 1M
$0.02
Cached / 1M
$0.02
Output / 1M
—

Parakeet TDT 0.6B v3

nvidia/parakeet-tdt-0.6b-v3

Usage
20 requests in the last 14 days
Context
—
Input / 1M
$16.67/1M s
Cached / 1M
—
Output / 1M
—

Gemma 4 E2B Instruct (real time)

google/gemma4-e2b-it:realtime

Usage
—
Context
—
Input / 1M
—
Cached / 1M
—
Output / 1M
—

31 models

Pricing calculator

Dirt-cheap Document OCR.

Pick a document type, page volume, and OCR model. See cost savings versus closed vision APIs.

Page volume / month

Gateway OCR tokens (in / out)

250M / 100M

Frontier billed tokens (in / out)

250M / 200M

2.5K image tokens per page; output includes reasoning at 2× OCR text.

Total costwhat the gateway bills for this workload

$45 - $343

Documents / month

100Kpages

Cost comparison (log scale)

Cheapest - most expensive
VLM Run Gateway
Document OCR VLMs
$45 - $343
OCR APIs
Textract, Azure Doc AI
$150 - $1K
Document AI APIs
Reducto, LlamaParse, Extend
$1K - $6K
Frontier VLMs
Gemini, Claude, GPT
$938 - $7.3K
$10$100$1K$10K

Estimates vs typical OCR, Document AI, and frontier VLM pricing.Source: llm-prices.com

A vision-only gateway, built for builders.

LLM routers and gateways route to 100s of LLMs, yet only a handful of VLMs, and often no OCR or classical CV models. Visual AI deserves its own stack.

  • Multimodal, Multitask

    One catalog spanning multi-modal inputs and multi-task outputs: OCR, detection, segmentation, pose, keypoints, and more.

  • Chat Completions Native

    OCR, captioning, and multimodal chat run through the same OpenAI-compatible chat completions API you already use.

  • Orchestration Built-In

    Send a 500-page PDF or a 2-hour video in a single call. The gateway chunks, batches, and reassembles for you. No pipelines to build.

  • Structured Outputs, Out of the Box

    JSON-schema enforcement on every call. CV wrappers emit fixed schemas; VLMs honor response_format.

  • Open Models, No Lock-In

    Every model is open-weight, served through the OpenAI-compatible API you already use. Swap models or providers freely, your client code never changes.

  • Agent-native vision, via MCP

    Give any MCP client instant access to the full visual model catalog. Agents see, read, and reason over images out of the box.

Built for production visual AI.

Optimized for production workloads and cost-efficiency. Every model deployed gets its own performance tune-up.

  • Models Supported

    30

  • p50 Latency

    <100ms

  • Uptime SLA

    99.9%

  • Type II · HIPAA · BAA

    SOC 2

12345678910111213141516171819
from openai import OpenAI

client = OpenAI(base_url="https://gateway.vlm.run/openai/",
                api_key="<VLM_RUN_API_KEY>")

resp = client.chat.completions.create(
    model="rednote-hilab/dots.mocr",
    messages=[{"role": "user", "content": [
        {"type": "document_url", "document_url": {"url": "invoice.pdf"}},
    ]}],
    extra_body={"method": "markdown"},
)
print(resp.choices[0].message.content)

# <document page="1">
# ## Invoice #1042
# | Item      | Qty | Amount |
# | API seats |  12 | $1,200 |
# </document>

Vision-native Chat Completions.

Point the OpenAI SDK at the gateway and swap the model, or hand the same catalog to an agent over MCP. Same signature, exhaustive visual model catalog.

Use cases

What you can build with the Gateway.

Compose vision models like building blocks. One SDK. One key. One bill.

Document ingestion pipelines

Layout, OCR, and markdown extraction across contracts, statements, and forms, at sub-cent per-page economics.

Multi-modal chatbots and agents

Build powerful multi-modal chatbots and agents with VLMs that pack native support for multi-modal inputs such as images, PDFs, or videos in a single chat completion.

Media and asset tagging

Auto-caption, tag, and index large image and video libraries for search, dedup, and recommendations.

Private and regulated deployments

Stand up the whole visual stack inside your own VPC or on-prem, with SSO, RBAC, and audit log export. Regulated data stays inside your boundary, under a BAA, on dedicated capacity you control.

Get an API key. Ship vision today.