# Compute Prices

> Compare cloud GPU and LLM inference API pricing across all major providers. Find the best rates for GPU compute and AI model APIs.

Compute Prices is a free, independent price comparison tool for cloud GPU rentals and LLM inference APIs. It aggregates real-time pricing data from 86 cloud providers and presents it in sortable, filterable tables so engineers and researchers can find the cheapest option for their workload.

The site is operated by Lansky Tech. It is not affiliated with any cloud provider.

## Key Pages

- [GPU Pricing](https://computeprices.com/gpu): Compare hourly GPU rental prices across providers. Filter by GPU model, provider, and pricing type (on-demand or spot).
- [LLM Inference Pricing](https://computeprices.com/inference): Compare per-token pricing for LLM inference APIs. Filter by model, provider, and modality.
- [LLM Models](https://computeprices.com/models): Browse all 347 tracked LLM models with specs, context windows, knowledge cutoff dates, and provider availability.
- [GPU Models](https://computeprices.com/gpus): Browse all tracked GPU models with specs, benchmarks, and current lowest prices.
- [Use Cases](https://computeprices.com/use-cases): Workload-shaped GPU guides covering LLM training, AI inference, fine-tuning, image generation, video rendering, scientific computing, RAG applications, and budget AI. Each page ranks suitable GPUs by suitability for that workload with live pricing context.
- [Providers](https://computeprices.com/providers): Directory of all cloud GPU and inference API providers with feature summaries.
- [Provider Comparison](https://computeprices.com/compare): Side-by-side comparison of any two providers on pricing, features, and GPU availability.
- [Learn](https://computeprices.com/learn): Educational content about cloud GPUs, pricing models, and AI infrastructure.
- [Blog](https://computeprices.com/blog): Articles on GPU pricing trends, provider updates, and cloud compute news.

## Data

Pricing data is collected automatically from provider websites and APIs every few hours. All GPU prices are normalized to per-GPU, per-hour USD. All LLM prices are normalized to per-1M-tokens USD. Prices shown are the most recent data point for each provider-model combination, kept separate per offering tier where a provider sells the same GPU in multiple variants (e.g. Runpod Secure Cloud vs Community Cloud) and per region where a provider charges different prices in different regions (e.g. AWS us-east-1 vs sa-east-1). The site shows the cheapest region per provider; full per-region prices are on /api/v1/gpu-prices.

Tracked GPU providers include AWS, CoreWeave, Lambda, Runpod, Vast.ai, Together AI, Hyperstack, FluidStack, Verda, GPU Outlet, and others.

Tracked LLM inference providers include OpenAI, Anthropic, Together AI, Groq, Fireworks AI, Mistral AI, AWS Bedrock, Google Vertex AI, Azure OpenAI, Replicate, DeepInfra, and others.

## API

A stable JSON API is available at `https://computeprices.com/api/v1`. Every endpoint requires an API key; a free one is instant (email magic link, no password) at https://computeprices.com/account/api-keys. Free keys get 750 requests/day, latest prices, the full catalog, weekly movers, and a 7-day price-trend window. The Pro subscription ($299/mo or $2,499/yr; 3,000/day, 3 keys) extends trends and raw price history to 90 days, plus upcoming price indices, provider comparison, and CSV/NDJSON export. Enterprise (custom throughput, 365-day windows) is available on request at api@computeprices.com.

- [API documentation](https://computeprices.com/docs/api): tier matrix, endpoint reference, authentication, rate limits, error model, roadmap.
- [OpenAPI 3.1 spec](https://computeprices.com/api/v1/openapi.json): machine-readable schema for every endpoint, no key needed. Use this to generate clients, register tools, or wire up an MCP server.
- Endpoints (all tiers): `/api/v1/gpu-prices`, `/api/v1/gpu-prices/history`, `/api/v1/llm-prices`, `/api/v1/gpus`, `/api/v1/gpus/{slug}`, `/api/v1/llm-models`, `/api/v1/llm-models/{slug}`, `/api/v1/providers`, `/api/v1/providers/{slug}`, `/api/v1/movers`, `/api/v1/trends/gpu/{slug}`, `/api/v1/trends/llm/{slug}`.
- Trend and history endpoints accept a `days` parameter clamped to the tier cap (free 7, pro 90, enterprise 365); `meta.tier_cap_days` reports the cap. `/gpu-prices/history` returns raw observations with the normalization audit trail (`original_price`, `normalization_method`) and paginates via `meta.next_cursor`.
- Responses include a `meta.upgrade_hint` field naming the next-tier features when relevant.

## Contact

- Website: https://computeprices.com
- LinkedIn: https://linkedin.com/in/gkobilansky
- GitHub: https://github.com/gkobilansky