AI Endpoints
InstantGPU AI Endpoints provide production-ready APIs for common AI workloads.
Use APIs to add AI capabilities to your application without managing the underlying model infrastructure yourself. InstantGPU hosts, optimizes, and auto-scales leading open-source models with dedicated vLLM and TensorRT acceleration so you can focus on building user-facing features.
Available AI Categories #
Choose from a comprehensive suite of pre-deployed, highly optimized AI models across all core modalities:
Generate, edit, transform, and upscale images using models like Flux.1, Stable Diffusion XL, and specialized LoRAs.
Generate and process AI video from text prompts, image-to-video pipelines, and high-framerate rendering engines.
Run speech recognition, fast transcription with Whisper-large-v3, and expressive text-to-speech workloads.
Use frontier language and generative AI models (Llama 3.3, Mistral, Qwen, DeepSeek) for text-based chat, reasoning, and coding.
Generate dense vector representations for semantic search, document retrieval, recommendation systems, and RAG knowledge-base applications.
Using an AI Endpoint #
The basic integration workflow consists of 4 direct steps:
Playground #
Use the InstantGPU Playground to test supported AI endpoints before integrating them into your application. Experiment with parameters, inspect raw JSON requests and responses, and generate code snippets in Python, cURL, and JavaScript directly from your browser.
Launch Interactive Playground
Test LLM prompts, diffusion models, and speech endpoints live.
Authentication #
All API requests require a valid InstantGPU API key passed via the standard HTTP Authorization header:
Authorization: Bearer YOUR_API_KEYMonitoring #
Use your dashboard to monitor endpoint activity, requests per minute, token latency, cold start times, error rates, and credit consumption in real time with detailed grafana-backed metrics.