# OpenGPU Radar

> Independent AI compute intelligence engine providing verified cloud GPU spot pricing, deterministic LLM VRAM calculations, and verified free LLM API directories. Built for engineers deploying open-source models at scale.

## Core Tools
- [LLM VRAM Calculator](https://opengpuradar.com/calculator): Deterministic memory sizing for weights, GQA KV cache, and runtime overhead. Supports FP16, FP8, INT4, and GGUF quantization modes.
- [Cloud GPU Spot Pricing](https://opengpuradar.com/cheapest-gpus): Daily verified spot and on-demand rental rates across 7+ providers. Filter by VRAM, price, and availability.
- [Free LLM API Radar](https://opengpuradar.com/free-llm): Verified free inference endpoints with rate limits, context windows, and no-card status. 11+ providers tracked.
- [AI Coding Config Hub](https://opengpuradar.com/free-llm/config): Config generator for Cursor, Claude Code, VS Code Continue, Cline, OpenCode, and Gemini CLI.
- [GPU Hardware Radar](https://opengpuradar.com/#gpus): Deep technical specs, break-even ROI calculators, and historical pricing for H100, H200, B200, A100, L40S, and RTX 4090.
- [Model Profiles](https://opengpuradar.com/models): Deployment profiles for Llama, DeepSeek, Qwen, Gemma, Mistral, Phi, and Command R with VRAM requirements.

## Guides
- [Cloud GPU Pricing Index](https://opengpuradar.com/guides/cloud-gpu-pricing-index): How to read and interpret spot vs on-demand pricing data.
- [Spot Instance Checkpointing](https://opengpuradar.com/guides/spot-instance-checkpointing): SIGTERM handlers, S3 checkpointing, and fault-tolerant training loops.
- [Free AI Coding Setup Guide](https://opengpuradar.com/guides/free-ai-coding-setup): Complete walkthrough for setting up Cursor, VS Code, and Gemini CLI with free LLM APIs.
- [Cursor Free Models Setup](https://opengpuradar.com/guides/cursor-free-models-setup): Step-by-step guide to configure Cursor with free models via OpenRouter and GitHub Models.

## Methodology & Research
- [Compute Data Methodology](https://opengpuradar.com/methodology): Mathematical formulas and pricing telemetry standards.
- [H100 vs H200 Long-Context Economics](https://opengpuradar.com/research/h100-vs-h200-long-context-economics): Sizing Llama 3.3 70B at 32k vs 128k context.
- [The KV Cache Memory Trap](https://opengpuradar.com/research/the-kv-cache-memory-trap): Why 70B models trigger CUDA OOM on dual 24GB GPUs.

## Provider Comparisons
- [GPU Comparison Engine](https://opengpuradar.com/compare): Factual memory bandwidth, interconnect, and cost comparisons (H100 vs H200 vs B200).
- [Compare Cloud GPU Providers](https://opengpuradar.com/compare/spheron-vs-runpod): Specialized cloud provider head-to-head analysis.

## API & Data
- Pricing data refreshed daily via automated GitHub Actions telemetry pipeline.
- All VRAM calculations use deterministic entity-graph formulas (not estimates).
- GPU pricing ground truth sourced from: Vast.ai, RunPod, Lambda Labs, Spheron, CoreWeave, Jarvis Labs, DataCrunch.