Features
- Spins up NVIDIA H100, A100, and newer GPU pods in under 30 seconds
- Routes jobs to the nearest node on a geo-distributed network
- Offers interruptible spot instances for batch processing and training
- Deploys models as serverless endpoints billed per millisecond
- Provides pre-configured environments for PyTorch, TensorFlow, JAX, LangChain, and AutoGPT
- Includes NVMe storage for fast dataset ingestion
- Supports SSH and JupyterLab access
Pricing Details
💵 Pricing (January 2026 Status) Medjed AI is known for its highly competitive, pay-as-you-go pricing model: Hardware Type Price (Approx.) Key Features NVIDIA A100 (80GB) ~$1.50 / hour High memory for large LLM training; Standard availability. NVIDIA H100 (NVL) ~$2.39 / hour The 2026 standard for high-speed inference and complex training. L40S / RTX 4090 ~$0.45 / hour Cost-effective for 3D rendering and light model fine-tuning. Serverless Inference Pay-per-token No hourly cost; Pay only for the compute used during the request. 🎁 Free Version / Trial: Medjed AI typically offers $10 - $25 in free credits for new users to test their GPU pods. They also provide a "Community Tier" with limited access to older-gen GPUs for open-source developers.