# ModelRunner

> ModelRunner is a developer platform for running AI models on the cloud through a unified API. It abstracts away provider complexity, exposing text-to-image, image-to-image, text-to-video, image-to-video and text-to-text (LLM) models from multiple providers (open-source via serverless GPU, Google, ByteDance, and more) behind a single REST API — plus an OpenAI-compatible chat completions endpoint for the language models — and a JavaScript SDK, with transparent, usage-based pricing.

ModelRunner gives developers granular control over model execution: full access to advanced schema parameters, model variants (speed vs. quality), transparent request lifecycles, and pricing modes including per-second, per-output, per-output-second, per-megapixel, resolution-tiered, megapixel-tiered, and per-token. Each model line below ends with `(category, price)`; `from $X` means the lowest rate that model can be charged at (a cheaper tier, or a cheaper option such as generating without audio), and `≈$X` is an average for models billed by GPU execution time.

## Product

- [Explore](https://modelrunner.ai/explore): Curated collections of AI models.
- [Models](https://modelrunner.ai/models): Browse and run open-source and provider AI models.
- [Wrappers](https://modelrunner.ai/wrappers): Custom model wrappers with prompt templates and base-model mappings.
- [Model creators](https://modelrunner.ai/owners): Every lab, company and author whose models run on ModelRunner, one page each.
- [Pricing](https://modelrunner.ai/pricing): Simple, usage-based pricing across all models.
- [Seedance 2.5](https://modelrunner.ai/seedance-2-5): Seedance 2.5 endpoints, specs, samples and pricing — 30-second single-take video with audio generated in the same pass, 480p/720p at 24 fps.
- [Serverless GPUs](https://modelrunner.ai/serverless-gpus): Deploy AI workloads on serverless GPUs — per-second billing, scale to zero, live GPU availability and rates, jobs submitted through the same queue API as catalog models.
- [Enterprise](https://modelrunner.ai/enterprise): ModelRunner for teams running AI in production — volume pricing, custom model onboarding, priority support, and a contact form that reaches the team directly (or email support@modelrunner.ai).
- [Open Source Program](https://modelrunner.ai/oss-program): Free API credits for open-source maintainers — a one-time credit from $50 (Starter) to $400–1,500+ (Flagship), tiered by project size and applied straight to the account balance; apply on the page or email support@modelrunner.ai.
- [fal.ai alternative](https://modelrunner.ai/alternatives/fal): How ModelRunner compares to fal.ai — one API key and one MCP server across providers, with published per-model rates.
- [Replicate alternative](https://modelrunner.ai/alternatives/replicate): How ModelRunner compares to Replicate — per-request published rates instead of per-second hardware billing, and one MCP server across providers.
- [Higgsfield alternative](https://modelrunner.ai/alternatives/higgsfield): How ModelRunner compares to Higgsfield for MCP users — 150+ models across providers behind one MCP server, billed per request instead of plan credits.

## Developers

- [Documentation](https://modelrunner.ai/docs): API references, integration guides, and tutorials.
- [MCP server](https://modelrunner.ai/mcp): Hosted Model Context Protocol server at https://mcp.modelrunner.run/mcp — exposes the public catalog to any MCP client as 23 callable tools over Streamable HTTP, authorized with OAuth 2.1 (no API key in your config). Setup for Claude Code, Claude Desktop and claude.ai, Cursor, and VS Code.
- [JavaScript SDK](https://github.com/modelrunner/modelrunner-js): Official client library for Node.js and browsers.
- [Sitemap](https://modelrunner.ai/sitemap.xml): Full list of indexable pages, including every public model page.
- [Full overview](https://modelrunner.ai/llms-full.txt): Long-form product reference for AI assistants.

## File uploads (Storage API)

Upload local files to ModelRunner storage and receive a temporary URL you can pass to model inputs (image, audio, video, or any binary file). All endpoints require an API key sent as `Authorization: Key <MRUN_API_KEY>`.

JavaScript SDK (Node.js, ESM):

```js
import { modelrunner } from "@modelrunner/client";
import fs from "node:fs";

const fileBuffer = fs.readFileSync("./image.jpeg");
const url = await modelrunner.storage.upload(new Blob([fileBuffer]));
console.log(url);

// Pass the returned url to any model input expecting a URL.
const result = await modelrunner.subscribe("swook/inspyrenet", {
  input: { image_path: url },
});
```

`Blob` is available natively in Node.js 18+. On older runtimes, upgrade or use a compatible polyfill.

cURL (two-step presigned upload):

```bash
# 1. Request a presigned upload URL
curl -X POST https://api.modelrunner.run/storage/upload/initiate \
  -H "Authorization: Key $MRUN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"file_name":"image.jpeg","content_type":"image/jpeg","size":12345}'
# → { "upload_url": "<presigned-s3-url>", "file_url": "https://media.modelrunner.ai/<key>" }

# 2. PUT the file bytes to upload_url (use the same content_type)
curl -X PUT "$UPLOAD_URL" \
  -H "Content-Type: image/jpeg" \
  --data-binary @./image.jpeg

# Use file_url as the input to your model run.
```

## Company

- [Home](https://modelrunner.ai/): Product overview.
- [Contact](mailto:support@modelrunner.ai): Email support.

## Legal

- [Terms of Service](https://modelrunner.ai/terms)
- [Privacy Policy](https://modelrunner.ai/privacy)

## Model creators

- [ByteDance](https://modelrunner.ai/owners/bytedance): 28 public models (image-to-image, image-to-video, text-to-image, text-to-video)
- [Google](https://modelrunner.ai/owners/google): 14 public models (image-to-image, image-to-video, music, sound)
- [Recraft](https://modelrunner.ai/owners/recraft): 11 public models (image-to-image, text-to-image)
- [roomix](https://modelrunner.ai/owners/roomix): 11 wrappers (image-to-image)
- [Wan Video](https://modelrunner.ai/owners/wan-video): 8 public models (image-to-video, text-to-video, video-to-video)
- [Bria](https://modelrunner.ai/owners/bria): 7 public models (image-to-image, video-to-video)
- [Kuaishou](https://modelrunner.ai/owners/kuaishou): 7 public models (image-to-video, text-to-video, video-to-video)
- [Black Forest Labs](https://modelrunner.ai/owners/black-forest-labs): 6 public models (image-to-image, text-to-image)
- [Alibaba](https://modelrunner.ai/owners/alibaba): 5 public models (image-to-video, text-to-image, text-to-text, text-to-video)
- [Resemble AI](https://modelrunner.ai/owners/resemble-ai): 5 public models (sound)
- [Stability AI](https://modelrunner.ai/owners/stability-ai): 5 public models (audio-to-audio, music, text-to-image)
- [ElevenLabs](https://modelrunner.ai/owners/elevenlabs): 4 public models (audio-to-audio, sound, speech-to-text)
- [HiDream.ai](https://modelrunner.ai/owners/hidream): 4 public models (image-to-image, text-to-image)
- [Microsoft](https://modelrunner.ai/owners/microsoft): 4 public models (image-to-image, image-to-text)
- [ModelRunner AI](https://modelrunner.ai/owners/modelrunner): 4 wrappers (image-to-image, text-to-image)
- [Tongyi-MAI](https://modelrunner.ai/owners/tongyi-mai): 4 public models (image-to-image, text-to-image)
- [Ideogram](https://modelrunner.ai/owners/ideogram): 3 public models (image-to-image, text-to-image)
- [Meta](https://modelrunner.ai/owners/meta): 3 public models (audio-to-audio, image-to-image, sound)
- [MiniMax](https://modelrunner.ai/owners/minimax): 3 public models (image-to-video, sound, text-to-video)
- [OpenAI](https://modelrunner.ai/owners/openai): 3 public models (image-to-image, speech-to-text, text-to-image)
- [Tencent](https://modelrunner.ai/owners/tencent): 3 public models (image-to-image, text-to-image)
- [xAI](https://modelrunner.ai/owners/xai): 3 public models (image-to-image, image-to-video, text-to-image)
- [Hyper3D](https://modelrunner.ai/owners/hyper3d): 2 public models (image-to-3d)
- [LongCat](https://modelrunner.ai/owners/meituan-longcat): 2 public models (image-to-video, text-to-video)
- [Luma](https://modelrunner.ai/owners/luma): 2 public models (text-to-image)
- [MEGVII Research](https://modelrunner.ai/owners/megvii-research): 2 public models (image-to-image, video-to-video)
- [Qwen](https://modelrunner.ai/owners/qwen): 2 public models (image-to-image, text-to-image)
- [TopazLabs](https://modelrunner.ai/owners/topazlabs): 2 public models (upscaler)
- [Tripo3D](https://modelrunner.ai/owners/tripo3d): 2 public models (image-to-3d, text-to-3d)
- [Z.ai](https://modelrunner.ai/owners/z-ai): 2 public models (text-to-text)
- [ACE Studio](https://modelrunner.ai/owners/ace-studio): 1 public model (music)
- [Boogu](https://modelrunner.ai/owners/boogu): 1 public model (text-to-image)
- [DeepSeek](https://modelrunner.ai/owners/deepseek): 1 public model (text-to-text)
- [Depth Anything](https://modelrunner.ai/owners/depth-anything): 1 public model (image-to-image)
- [FASHN](https://modelrunner.ai/owners/fashn): 1 public model (image-to-image)
- [FireRedTeam](https://modelrunner.ai/owners/fireredteam): 1 public model (image-to-image)
- [grey-hound432](https://modelrunner.ai/owners/grey-hound432): 1 public model (image-to-image)
- [Hendrik Schröter](https://modelrunner.ai/owners/rikorose): 1 public model (audio-to-audio)
- [hexgrad](https://modelrunner.ai/owners/hexgrad): 1 public model (sound)
- [Krea](https://modelrunner.ai/owners/krea): 1 public model (text-to-image)
- [Lightricks](https://modelrunner.ai/owners/lightricks): 1 public model (sound)
- [lllyasviel](https://modelrunner.ai/owners/lllyasviel): 1 public model (image-to-image)
- [MMAudio](https://modelrunner.ai/owners/mmaudio): 1 public model (video-to-video)
- [Moondream](https://modelrunner.ai/owners/moondream): 1 public model (image-to-text)
- [NightmareAI](https://modelrunner.ai/owners/nightmareai): 1 public model (upscaler)
- [Peng Zheng](https://modelrunner.ai/owners/zhengpeng7): 1 public model (image-to-image)
- [philz1337x](https://modelrunner.ai/owners/philz1337x): 1 public model (upscaler)
- [PixVerse](https://modelrunner.ai/owners/pixverse): 1 public model (text-to-video)
- [shallowdream204](https://modelrunner.ai/owners/shallowdream204): 1 public model (text-to-image)
- [Shangchen Zhou](https://modelrunner.ai/owners/sczhou): 1 public model (image-to-image)
- [StepFun AI](https://modelrunner.ai/owners/stepfun-ai): 1 public model (image-to-text)
- [swook](https://modelrunner.ai/owners/swook): 1 public model (image-to-image)
- [Sync](https://modelrunner.ai/owners/sync): 1 public model (video-to-video)
- [TencentARC](https://modelrunner.ai/owners/tencentarc): 1 public model (image-to-image)

## Blog

- [AI Credits for Open Source Maintainers: 5 Programs Compared](https://modelrunner.ai/blog/ai-credits-for-open-source-maintainers): AI credits for open-source maintainers, compared: what each 2026 program grants, who qualifies, and how to pick — OpenAI, Anthropic, Ona, Atlas Cloud.
- [Image Generation API Pricing in 2026: 15 Models Compared](https://modelrunner.ai/blog/image-generation-api-pricing): Image generation API pricing compared across 15 text-to-image models, from $0.0025 to $0.25 per image. Live per-model rates, pay-per-use, no subscription.
- [Turn Property Photos into a Real Estate Video with AI (for About $1)](https://modelrunner.ai/blog/real-estate-video-from-photos-with-ai): Turn listing photos into a 30-second real estate video with AI image-to-video. Step-by-step workflow, camera prompts, and real pricing: $0.90 vs $5.43.
- [Higgsfield Alternative for MCP Users](https://modelrunner.ai/blog/higgsfield-alternative-for-mcp-users): Higgsfield's MCP server is locked to its own credits and roster. ModelRunner's MCP gives your AI assistant 150+ models across providers, pay-per-use.
- [How to Get 4K, High-Quality AI Images Without Overpaying](https://modelrunner.ai/blog/high-quality-4k-ai-images-the-cheap-way): Get 4K, high-quality AI images without overpaying: generate on dedicated models, upscale only the keepers, and explore at low resolution to cut cost.
- [Generating a Vertical Video Ad With AI: Keyframe Anchoring](https://modelrunner.ai/blog/generating-a-vertical-video-ad-with-ai-keyframe-anchoring): Generate a vertical video ad with AI for an iOS app: keyframe anchoring lets an image model own product identity and a video model own only the motion.
- [Generating App Store Screenshots With AI: A Home Redesign App Case Study](https://modelrunner.ai/blog/generating-app-store-screenshots-with-ai-home-redesign-app-case-study): Generate App Store marketing screenshots with AI for a home redesign iOS app, device mockups and ASO-aware headlines included, in one Claude Code session.
- [Generating Demo Assets for an iOS App With AI: A Home Redesign App Case Study](https://modelrunner.ai/blog/generating-demo-assets-for-an-ios-app-with-ai): Claude Code has no image generation on its own — connect the ModelRunner MCP and it calls image models as native tools. A real 29-image iOS batch shows how.

## Featured Models

- [Seedance 2.5 Image to Video](https://modelrunner.ai/models/bytedance/seedance-v2.5/image-to-video): Animate a still photo into a video with synchronized audio — single takes up to 30 seconds long that keep your image's exact shape. (image-to-video, from $0.154 per second of video)
- [Happy Horse 1.1 Text to Video](https://modelrunner.ai/models/alibaba/happy-horse/v1.1/text-to-video): Generate a short video with synchronized native audio from a text prompt, with spoken dialogue lip-synced on screen, at 720P or 1080P. (text-to-video, from $0.14 per second of video)
- [Lyria 3 Clip](https://modelrunner.ai/models/google/lyria-3/clip): Generate a 30-second song clip — vocals and lyrics included by default — from a single text prompt, as an MP3. (music, $0.04 per audio clip)
- [Nano Banana 2 Lite Text to Image](https://modelrunner.ai/models/google/nano-banana-2-lite): Generate 1K (roughly 1 megapixel) images from a text prompt at the lowest price in the Nano Banana family, with 14 aspect ratios from square to ultra-wide banner. (text-to-image, $0.034 per image)
- [Kling 3.0 Motion Control](https://modelrunner.ai/models/kuaishou/kling-video/v3/motion-control): Transfer motion and facial expression from a driving video onto a character image, animating your character to perform the reference movements. (video-to-video, $0.168 per second of video)
- [Seedream 5.0 Pro Text to Image](https://modelrunner.ai/models/bytedance/seedream-v5-pro/text-to-image): Generate high-resolution images from a text prompt, with accurate in-image text and crisp detail — the premium tier of Seedream 5.0. (text-to-image, from $0.045 per image)
- [Stable Audio 2.5 Audio-to-Audio](https://modelrunner.ai/models/stability-ai/stable-audio-2.5/audio-to-audio): Transform an existing audio clip into new music or sound effects guided by a text prompt — restyle, re-instrument, or reimagine a source track, returned as a WAV. (audio-to-audio, $0.2 per audio clip)
- [Recraft V4.1 Pro](https://modelrunner.ai/models/recraft/v4.1/pro/text-to-image): Generate a polished, high-resolution raster image from a text prompt — Recraft's newest Pro tier, producing imagery up to 2048×2048 with professional composition, lighting, and materials. (text-to-image, $0.21 per image)

## Language models (OpenAI-compatible chat completions)

Text-to-text models are served by a synchronous OpenAI-compatible endpoint at `POST https://queue.modelrunner.run/{owner}/{alias}/chat/completions` — NOT by the asynchronous queue API below (the queue submit returns 400 for them). Point any OpenAI SDK or OpenAI-compatible tool at `base_url = https://queue.modelrunner.run/{owner}/{alias}` with a ModelRunner API key (`Authorization: Bearer <key>`); `model` is the same `{owner}/{alias}`. Streaming (SSE), tool calling, JSON mode and reasoning effort work where the model supports them; each model's own `llms.txt` carries its parameters and limits. Rates are per 1M tokens.

- [Qwen3.8-Max](https://modelrunner.ai/models/alibaba/qwen3.8-max): Answer questions about images, video and text in one chat model, with thinking effort adjustable from fully off to maximum and a 1,000,000-token context. ($2 in · $6 out per 1M tokens, 1M context)
- [GLM-5.2 Fast Preview](https://modelrunner.ai/models/z-ai/glm-5.2-fast-preview): Low-latency chat completions from the GLM-5.2 weights on throughput-tuned serving — consistently faster than the standard tier, at twice the price and identical answer quality. ($2.8 in · $8.8 out per 1M tokens, 1M context)
- [DeepSeek V4 Pro](https://modelrunner.ai/models/deepseek/v4): Open-weight (MIT) thinking model at 1.6T parameters for hard reasoning and competitive-grade coding, with a 1,000,000-token context, tool calling and streaming over an OpenAI-compatible chat endpoint. ($2.4 in · $4.8 out per 1M tokens, 1M context)
- [GLM-5.2](https://modelrunner.ai/models/z-ai/glm-5.2): Open-weight (MIT) thinking model for agentic coding and long-horizon reasoning, with a 1M-token context, seven levels of thinking effort, tool calling and streaming over an OpenAI-compatible chat endpoint. ($1.4 in · $4.4 out per 1M tokens, 1M context)
- [Gemini 3.5 Flash-Lite](https://modelrunner.ai/models/google/gemini-3.5-flash-lite): The cheapest Gemini text tier — built for high-volume agentic tasks, translation and simple data processing over an OpenAI-compatible endpoint. ($0.3 in · $2.5 out per 1M tokens, 1M context)
- [Gemini 3.7 Flash](https://modelrunner.ai/models/google/gemini-3.7-flash): Reasoning-first flagship text model for coding and agentic work, with tunable thinking levels, tool calling and streaming over an OpenAI-compatible endpoint. ($0.75 in · $3.75 out per 1M tokens, 1M context)
- [Gemini 3.5 Flash](https://modelrunner.ai/models/google/gemini-3.5-flash): Fast, general-purpose text model served over an OpenAI-compatible chat completions endpoint, with tool calling, JSON mode and streaming. ($1.5 in · $9 out per 1M tokens, 1M context)