Prime Intellect
Compute aggregation, inference, and RL training
Prime Intellect is an AI research company and compute platform. It aggregates on-demand GPUs and reserved clusters from many providers behind one account, and runs Prime Intellect Inference, an OpenAI-compatible API that serves open-weight models directly and routes to proprietary models, plus hosted reinforcement-learning training through its Lab platform.
At a glance · prices collected Nov 12, 2026
What's good about Prime Intellect
- One account for GPUs across many underlying providers
- Per-token inference rates are published through a public models endpoint
- Inference, training, and GPU rental on the same platform
- Publishes open-source models, datasets, and RL tooling
Prime Intellect LLM API prices
Pay-per-token pricing, per 1M tokens.
Prices last updated November 12, 2026
| Model | Input/1M | Output/1M | ||||
|---|---|---|---|---|---|---|
| $0.027 | $0.201 | |||||
| $0.039 | $1.20 | |||||
| $0.050 | $0.400 | |||||
| $0.051 | $0.335 | |||||
| $0.052 | $0.250 | |||||
| $0.060 | $0.240 | |||||
| $0.070 | $0.200 | |||||
| $0.070 | $0.400 | |||||
| $0.100 | $0.400 | |||||
| $0.100 | $0.300 | |||||
Prime Intellect pricing by model
Input, output, and batch rates, plus alternatives, for one model at a time.
Billing & terms
- On-demand GPUs
- Hourly rates set per underlying provider, with community and spot tiers where available
- Reserved clusters
- Quoted per request from participating data centers
- Per token
- Inference billed per 1M input and output tokens, with cached-input rates on hosted models
- Support
- Documentation, plus direct assistance from the research and infrastructure engineering team and a dedicated solutions engineer for reserved-cluster customers
Watch out for
- GPU rates are only available after sign-in through the availability API, so they are not tracked here
- As an aggregator, hardware, networking, and support depend on the underlying provider
- Gateway models are routed to other inference providers rather than served on Prime Intellect's own hardware
Features
- Multi-provider GPU access
- On-demand access to 1 to 256 GPUs sourced from many clouds, managed from one dashboard, CLI, and API
- Reserved clusters
- Large-cluster requests sent to 50+ data centers for parallel quotes, from H100 and H200 to B200, B300, and GB300 NVL72
- Prime Intellect Inference
- OpenAI-compatible API for hosted open-weight models and gateway access to OpenAI, Anthropic, Google, xAI, and other model families, billed per token
- Hosted RL training
- Lab runs reinforcement-learning post-training against environments from the Environments Hub, with LoRA adapters deployable for inference
- Orchestration and networking
- SLURM and Kubernetes orchestration, InfiniBand networking, and Grafana monitoring for clusters
- Spot resale
- Idle reserved nodes can be put on Prime Intellect's spot market
Company
Published by Prime Intellect to identify the company behind this listing.
- Headquarters
- 🇺🇸 San Francisco, United States
Getting started
Step 1
Create an account
Sign up at app.primeintellect.ai
Step 2
Rent GPUs
Pick a GPU type, count, and provider from the availability list, then provision from the dashboard or the prime CLI
Step 3
Call the inference API
Create an API key and point an OpenAI-compatible client at api.pinference.ai
Where this page's facts come from
- LLM API prices
- docs.primeintellect.ai · last collected Nov 12, 2026
- Profile
- Reviewed Sep 28, 2026
- Method
- How we collect and check prices
Frequently Asked Questions
Which models does Prime Intellect serve?
Check the pricing table above for the models Prime Intellect currently serves and their per-token rates.
How do I get started with Prime Intellect?
Create an account, Rent GPUs, Call the inference API
What are Prime Intellect's main advantages?
Prime Intellect's main advantages include: One account for GPUs across many underlying providers, Per-token inference rates are published through a public models endpoint, Inference, training, and GPU rental on the same platform, Publishes open-source models, datasets, and RL tooling.
What are Prime Intellect's limitations?
Prime Intellect's main limitations include: GPU rates are only available after sign-in through the availability API, so they are not tracked here, As an aggregator, hardware, networking, and support depend on the underlying provider, Gateway models are routed to other inference providers rather than served on Prime Intellect's own hardware.
Compare Providers
Find the best prices for the same GPUs and models from other providers · All Prime Intellect alternatives, ranked →
All Prime Intellect Comparisons
The weekly price report
One email every Monday with the week's notable GPU and LLM API price moves, plus occasional product news. No spam.