Rank 62 AI inference providers across 61 companies—from wafer-scale ASICs (Cerebras, Groq, SambaNova, Taalas) and serverless GPU clouds (Together, Fireworks, Baseten, DeepInfra) to hyperscalers, frontier lab APIs, sovereign European clouds, and media/voice platforms. Compare speed-versus-price Pareto frontiers, run a live token streaming race, trace accelerator silicon, or map data-center residency below.
0 of 25 ranked
Drop providers here
Drop providers here
Drop providers here
Drop providers here
Drop providers here
Drop providers here
Drag providers into tiers, or tap one and then a tier. Keyboard: focus a provider, press 1–6 to rank it and 0 to
unrank, or Enter then the arrow keys to move it. Ctrl+Z undoes.
Unranked
Type
Silicon
Modality
Precision
Features
Data centers
Origin
Company
Best for
Sort
Popular includes the most widely used inference clouds and frontier APIs tracked by Artificial Analysis (benchmarks, tracked LLM hosts), and BenchLeader (frontier labs, top-50 model APIs). Switch to Open-Weight Hosts or All for 37 more.
Quantization
Speed × Price Frontier
Throughput (tok/s) versus blended $/1M token cost and the Pareto efficiency curve.
On gpt-oss-120b, Cerebras Inference leads throughput at 759 tok/s ($0.45/1M),
while Venice AI is the
lowest-cost host at $0.060/1M (42 tok/s). 7 hosts sit on the speed-per-dollar Pareto frontier.