llms.txt
# Runware
> Runware is AI infrastructure rebuilt to cost less: one API for 400K+ AI models, serverless GPUs and reserved GPU capacity, all on modular data centers Runware designs, builds and runs. Usage-based pricing, half the cost of a neocloud.
Runware runs on its own Sonic Inference Engine® and Sonic Pods, modular AI data centers built for training and inference. Customers can use the Models API (400K+ hosted models behind one API), Serverless (their own code or containers on Runware GPUs), Compute (reserved GPU nodes, clusters or serverless capacity on 2 to 4 year terms) or Model Serving (their own model weights served as a production API). Pricing is usage-based with free test credits ($2 with a business email signup, no credit card required).
## Products
- [Models API](https://runware.ai/models-api): one AI API for image, video, audio, LLM and 3D models. Open and proprietary models with the same auth and billing, a JSON schema per model, up to 10x below market rates.
- [Serverless](https://runware.ai/serverless): serverless GPUs for any AI workload. Run any container or model with fast cold starts, scale to zero and per-second billing, for inference, training and fine-tuning. Billed per second from $1.60 per GPU-hour on demand, or reserve capacity from $0.63 per GPU-hour.
- [Compute](https://runware.ai/compute): reserved GPU capacity for AI training and inference on 2 to 4 year terms, as bare-metal nodes and clusters or guaranteed serverless capacity on Sonic Pods, priced per GPU-hour. Half the cost of a neocloud. HGX B300, GB300 NVL72, RTX PRO 6000 and Vera Rubin NVL72 allocating now.
- [Model Serving](https://runware.ai/model-serving): send Runware your model weights and they are served as a production API, public or private. No containers to manage, served from a shared model lake across pods for built-in redundancy, billed per inference.
- [Pricing](https://runware.ai/pricing): pay-as-you-go pricing for AI models (per generation or per token), serverless GPUs billed by the second and reserved GPU capacity. No subscriptions.
## Models
- [Model directory](https://runware.ai/models): browseable catalog of every model available through the API. By modality: [LLMs](https://runware.ai/models?io=text-to-text), [Image models](https://runware.ai/models?io=text-to-image), [Video models](https://runware.ai/models?io=text-to-video), [Audio models](https://runware.ai/models?io=text-to-audio), [3D models](https://runware.ai/models?io=text-to-3d).
- [Collections](https://runware.ai/collections): curated groupings of models by use case.
Modalities and model counts in the Models API:
- Image: 211 models
- Video: 106 models
- Audio: 29 models
- Text (LLMs): 55 models
- Vision: 51 models
- Editing: 91 models
- Upscale: 15 models
- BG removal: 13 models
- 3D: 7 models
For schema-level reference of every task type and parameter, see [docs/llms.txt](https://runware.ai/docs/llms.txt).
## Features
- [Image Generation API](https://runware.ai/image-generation-api): lowest-cost image generation API. Text-to-image, image-to-image, image editing, inpainting, outpainting, upscaling, background removal and image-to-3D on 400K+ models, with LoRA, ControlNet and IP-Adapter support.
- [Video Generation API](https://runware.ai/video-generation-api): text-to-video, image-to-video, video-to-video and lip sync on one endpoint, pay per video. Veo, Kling, Seedance, Wan and more.
- [Audio Generation API](https://runware.ai/audio-generation-api): text-to-speech, voice cloning, music and sound effects through one audio API. MiniMax, Gemini TTS, Fish Audio and more, pay as you go.
- [LLM Inference](https://runware.ai/llm-api): LLM inference on one API, two ways. The on-demand Model API serves open-weight and proprietary language models, billed per token with no subscription, in the OpenAI Chat Completions format and, for Claude models, the Anthropic Messages format. Dedicated LLM inference reserves capacity for a production workload, with guaranteed throughput, increased rate limits, volume-based pricing, custom SLAs, custom models and region-locked data processing. Endpoint: https://api.runware.ai/v1/chat/completions
- [LoRA Training](https://runware.ai/lora-training): train a style LoRA from your own images, then generate with it on the same API. Private by default, weights downloadable anytime.
- [Model Upload](https://runware.ai/docs/platform/model-upload): bring your own checkpoints, LoRAs and ControlNets to the Models API.
- [Enterprise Managed Platform](https://runware.ai/enterprise-managed-platform): managed generative AI platform for enterprise teams, with access controls, monitoring, data exports and direct technical support.
## Tools and integrations
- [Runware MCP server](https://runware.ai/mcp): connect Claude, Cursor, Codex or any MCP client to every Runware model with one URL. Setup guide: https://runware.ai/docs/platform/mcp
- [Runware CLI](https://runware.ai/cli): a single static binary to run any model from the terminal, plus search, presets, uploads, account management and Serverless deploys. Docs: https://runware.ai/docs/platform/cli
- [Runware Skills](https://runware.ai/skills): 30+ open-source agent skills that teach coding agents to get the best results from every Runware model across image, video, audio, 3D and text. Docs: https://runware.ai/docs/platform/skills
- [Migrate to Runware](https://runware.ai/migrate-to-runware): guidance for moving from fal, Replicate or WaveSpeed. Same request shape, one API for every model, plus free migration Skills and hands-on help.
- [OpenClaw](https://runware.ai/free-openclaw): free hosted OpenClaw agent on a secure VPS, wired to Runware models, pay per token.
- [Playground](https://runware.ai/playground): run any model in the browser.
- [JavaScript SDK](https://www.npmjs.com/package/@runware/sdk-js) and [Python SDK](https://pypi.org/project/runware/): official SDKs.
- [ComfyUI](https://runware.ai/docs/platform/comfyui) and [Vercel AI SDK](https://runware.ai/docs/platform/vercel-ai): integrations for running Runware models from existing tooling.
## Developers
- [Documentation](https://runware.ai/docs): API reference, SDK guides, task types and model parameters.
- [Models API docs](https://runware.ai/docs/models-api/introduction): one request shape for every modality, authentication, pricing, webhooks, polling and errors.
- [Serverless docs](https://runware.ai/docs/serverless/introduction): quickstart, writing a model, bringing a container, compute and scaling, volumes, pricing and migration guides.
- [OpenAI-compatible APIs](https://runware.ai/docs/platform/compatible-apis): using Runware through OpenAI-compatible clients.
- [docs/llms.txt](https://runware.ai/docs/llms.txt): machine-readable index of the docs site for LLM consumption.
- [Guides](https://runware.ai/docs/guides): tutorials and how-tos.
- [GitHub](https://github.com/runware): open-source SDKs, CLI, Skills and examples.
- [Status](https://status.runware.ai/): platform status and incident history.
- [Discord](https://discord.gg/runware): community and support.
## Technology
- [Sonic Inference Engine®](https://runware.ai/sonic-inference): Runware's AI inference platform, built from the hardware up. Sonic Pods hardware and Sonic Inference software engineered as one system: sub-second cold starts, a shared model lake of 400K+ models, up to 10x lower cost.
- [Sonic Pods](https://runware.ai/sonic-pods): low-cost modular AI data centers. Up to 1,200 directly liquid-cooled GPUs in a 20 ft container, running any GPU architecture for training or inference. Built for less, live in days.
## Resources
- [Blog](https://runware.ai/blog): product announcements, model launches and engineering posts.
- [Case studies](https://runware.ai/case-studies): how AI teams cut costs, launch new models in hours and scale on Runware.
- [AI API price comparison](https://runware.ai/alternative): Runware model pricing against other providers.
- [fal alternative](https://runware.ai/fal-alternative), [Replicate alternative](https://runware.ai/replicate-alternative), [WaveSpeed alternative](https://runware.ai/wavespeed-alternative): side-by-side comparisons.
- [RunPod alternative](https://runware.ai/runpod-alternative): Runware Serverless GPU rates against RunPod Serverless, per GPU-hour, pay as you go and reserved.
- [Modal alternative](https://runware.ai/modal-alternative): Runware Serverless GPU rates against Modal, per GPU-hour, Serverless and reserved.
- [DeepInfra alternative](https://runware.ai/deepinfra-alternative): Runware LLM token rates and FLUX image rates against DeepInfra, model by model.
- [Baseten alternative](https://runware.ai/baseten-alternative): Runware Serverless GPU rates against Baseten dedicated deployments, per GPU-hour, then open LLM token rates against Baseten Model APIs.
- [Together AI alternative](https://runware.ai/together-ai-alternative): Runware open-LLM token rates, per-image rates and Serverless GPU-hour rates against Together AI, model by model and GPU by GPU.
- [Fireworks AI alternative](https://runware.ai/fireworks-ai-alternative): Runware Serverless GPU rates against Fireworks AI on-demand deployments, per GPU-hour, then open-LLM token rates against Fireworks AI serverless inference, including models it serves on dedicated deployments only.
## Company
- [About](https://runware.ai/about): Runware builds AI infrastructure from the hardware up. Story, team and how we work.
- [Contact](https://runware.ai/contact): support, sales, enterprise, reserved capacity and dedicated infrastructure enquiries.
- [Creator Program](https://runware.ai/creator-program): monthly credits for creators publishing work built on Runware.
- [Startup Program](https://runware.ai/startup-program): up to $30,000 in credits for AI startups building on Runware. Growth: deposit $25,000 and Runware adds $10,000 in credits (40%). Scale: deposit $50,000 and Runware adds $30,000 (60%). Starter: up to $1,000 in free credits for bootstrapped and pre-seed teams. Credits do not expire and work on every model through one API.
- [Trust](https://runware.ai/trust): security and compliance posture. SOC 2 and ISO 27001 certified, GDPR compliant, EU and US data residency, SSO and SAML.
## Policies
- [Terms](https://runware.ai/terms)
- [Privacy](https://runware.ai/privacy)
- [Cookies policy](https://runware.ai/cookies-policy)
- [Trust & abuse reporting](https://runware.ai/trust)
- [Security disclosure](https://runware.ai/security-disclosure)
