State of inference
State of inference will compare task results and costs to help you choose an AI model, decide where to run it and know when to switch.
Research space
Read moreA rotating point cloud with labels for models, providers, workflows and economics. An abstract illustration of the research, not measured data.
State of inference
State of Inference brings model lifecycle records, provider comparisons, reference workflows and economic models into one public research terminal. It connects changes in models and providers to the cost and reliability of complete workflows. The datasets preserve sources, measurement conditions and earlier versions so findings can be inspected and comparisons revisited.
4 research areasModels, providers, workflows and economics.
Register your interest in future updates. Keep me updated
State of Inference will bring together our research on models, providers, workflows and economics. The aim is to connect changes in the AI market to their effect on useful work, so businesses can judge whether a new model or a different setup is worth adopting.
Comparisons will show what was tested, how well it completed the task and what it cost. Methods, sources and earlier results will remain available so you can inspect the evidence behind a conclusion.
State of Inference is not yet available. Register your interest for updates.