290+ models · 40+ providers · real-time pricing

Choose the right inference
for your customer support

Compare LLM providers. Size GPUs for self-hosting.
Understand the economics behind every token.

Compare LLM Pricing→Estimate VRAM
Tools

Three ways to reason about AI costs

B2B presets
01

Choose an LLM Provider

Pick your use case — customer support, RAG, code gen — and compare per-call and monthly costs across all providers.

System prompt + user context split · prompt caching estimates

02

Size GPUs for Your Model

Estimate VRAM for any model — weights, KV-cache, precision, safety margins. Find the right accelerator.

A100, H100, L40S, 4090 · side-by-side comparison

03

Understand Inference Economics

How caching, batching, quantization, and routing affect your real spend. Data center trends and supply analytics.

Cost optimization · industry benchmarks · hardware supply

04

Optimization Playground

Experiment with prompt caching, model routing, and quantization strategies. See cost impact before you ship.

05

Live Model Playground

Test models side-by-side. Compare response quality, latency, and cost in real-time.

What you can do
Compare inference costs for your exact workload
Estimate VRAM requirements for inference and training
Compare accelerators — real memory and bandwidth
Understand trade-offs: context, batch size, precision
Connect model choices to cost and scalability
Philosophy

A technical planning tool for people who design, deploy, and scale LLM systems. Focused on how models actually consume memory and bandwidth.

No benchmarks for marketing. No abstract performance scores.
All assumptions are explicit and inspectable.

ML EngineersCTOsPlatform TeamsFounders