# PlanetGPU — llms.txt

## Description
PlanetGPU is a multi-cloud GPU price and specification index. It aggregates
live GPU pricing from 12 cloud providers and pairs those offers with curated
GPU specs from GPU Ark: VRAM, FP16 throughput, memory bandwidth, TDP,
architecture, cores, clocks, tensor cores, and release data. Model VRAM fit and
deployment helpers are available as secondary workflows. Data is powered by
gpuhunt (pricing), GPU Ark (specs), Hugging Face metadata, and a model/GPU
knowledge graph.

## Core Tasks
- Compare GPU cloud prices across AWS, GCP, Azure, RunPod, Vast.ai, Lambda Labs,
  Nebius, OCI, Vultr, CloudRift, Cudo Compute, and Verda.
- Browse and filter GPU specifications including VRAM, bandwidth, FP16/FP32
  throughput, architecture, tensor cores, TDP, and release data.
- Compare two GPUs by live price floors, provider coverage, and core specs.
- Compare FP16 and INT4 VRAM requirements for LLMs.
- Recommend the cheapest GPU + provider to run a given model within a budget.
- Estimate monthly cost-to-serve for an LLM workload.
- Score GPUs for specific AI workloads (training, inference, fine-tuning, edge).
- Generate deployable dstack configs for a model on a chosen GPU.

## Pages
- Home: https://planetgpu.com
- GPU Search: https://planetgpu.com/search
- GPU Detail: https://planetgpu.com/search/{gpu}
- Browse GPUs: https://planetgpu.com/gpus
- Leaderboard: https://planetgpu.com/leaderboard
- Workload Scorer: https://planetgpu.com/score
- Find a GPU: https://planetgpu.com/find
- Efficiency Rankings: https://planetgpu.com/efficiency
- Budget Planner: https://planetgpu.com/budget
- Cluster Planner: https://planetgpu.com/cluster
- GPU Timeline: https://planetgpu.com/timeline
- Models: https://planetgpu.com/models
- Model Detail: https://planetgpu.com/model/{slug}
- Frameworks: https://planetgpu.com/framework/{slug}
- Providers: https://planetgpu.com/provider/{slug}
- Compare: https://planetgpu.com/compare/{a}-vs-{b}
- Recipes: https://planetgpu.com/recipes
- Deploy: https://planetgpu.com/deploy
- Calculator: https://planetgpu.com/calculator
- Price Index: https://planetgpu.com/index/{gpu}
- Wiki: https://planetgpu.com/wiki
- Docs: https://planetgpu.com/docs

## API
Base URL: https://api.planetgpu.com/v1

### Endpoints
- GET /v1/search?level=gpu — Aggregated GPU cards across providers
- GET /v1/search?level=provider&gpu={gpu} — Provider comparison for a GPU
- GET /v1/search?level=offer&gpu={gpu}&provider={p} — Individual SKU offers
- GET /v1/leaderboard/models — Model VRAM requirements and Hugging Face freshness
- POST /v1/recommend — Recommend GPU + provider + cost + dstack config for a model
- POST /v1/reason/deploy — Deployment reasoning for a goal (users, budget, region)
- POST /v1/repos/analyze — Analyze a repo and get deployment recommendations
- POST /v1/calculator/cost-to-serve — Cost-to-serve calculator
- GET /v1/compare?a={gpu}&b={gpu} — Side-by-side GPU comparison
- GET /v1/index/{gpu} — Public price index for a GPU
- GET /v1/gpu-specs — GPU specifications (powered by GPU Ark)
- GET /v1/graph/nodes?node_type={type}&limit={n} — Knowledge graph nodes (paginated)
- GET /v1/graph/recommend/recipe?model_id={id} — Recommended recipe for a model
- GET /v1/wiki/articles — Compiled wiki articles
- GET /v1/recipes — Deployment recipe gallery

### Example: recommend a deployment
POST /v1/recommend
{"model_id": "llama-3-70b", "workload": "inference", "monthly_budget_usd": 3000}

## Attribution
GPU specifications are powered by GPU Ark. Pricing is aggregated via gpuhunt.
Please retain attribution when reusing specification or pricing data.