---

name: prime-intellect description: Prime Intellect GPU cloud API reference. Use when working with Prime Intellect pods, GPUs, clusters, disks, SSH keys, inference, or any PI compute resources.

Prime Intellect API Reference

Base URL: https://api.primeintellect.ai Inference API: https://api.pinference.ai/api/v1

All requests require Authorization: Bearer <api_key> header. For team accounts, include X-Prime-Team-ID: <team_id> header.

API Groups

• Availability - Check GPU, cluster, and disk availability and pricing
• Pods - Create, manage, and monitor GPU instances
• Disks - Network-attached storage management
• SSH Keys - Manage SSH keys for pod access
• Inference - Models and chat completions API
• Sandboxes - Lightweight container environments
• Images - Custom image management
• Evals - Evaluation management
• User - User info and teams

Quick Reference

Available GPU Types

CPU_NODE, A10_24GB, A100_80GB, A100_40GB, A30_24GB, A40_48GB, B200_180GB, B300_262GB, RTX3090_24GB, RTX4090_24GB, RTX5090_32GB, H100_80GB, H200_96GB, GH200_96GB, L4_24GB, L40_48GB, L40S_48GB, A4000_16GB, A5000_24GB, A6000_48GB, V100_16GB, V100_32GB, T4_16GB

Available Regions

africa, asia_south, asia_northeast, australia, canada, eu_east, eu_north, eu_west, middle_east, south_america, united_states

Available Images

ubuntu_22_cuda_12, cuda_12_1_pytorch_2_2, cuda_11_8_pytorch_2_1, cuda_12_1_pytorch_2_3, cuda_12_1_pytorch_2_4, cuda_12_4_pytorch_2_4, cuda_12_4_pytorch_2_5, cuda_12_4_pytorch_2_6, cuda_12_6_pytorch_2_7, stable_diffusion, axolotl, bittensor, vllm_llama_8b, vllm_llama_70b, vllm_llama_405b, custom_template, flux, prime_rl

Provider Types

runpod, fluidstack, lambdalabs, hyperstack, oblivus, cudocompute, scaleway, tensordock, datacrunch, latitude, crusoecloud, massedcompute, akash, primeintellect, primecompute, nebius, vultr, denvr

Socket Types

PCIe, SXM2, SXM3, SXM4, SXM5, SXM6

Security Types

secure_cloud, community_cloud