0

Choose your Instrument

Pick the right intelligence tier for your task, or route across them for peak performance at lower costs.

From zero to first request
in < 5 minutes

Every AI task. Covered.

1Prices are refreshed frequently, but they move in real time with the market. For the latest price, visit The Grid App.

2The qualifying model list changes as new models meet the instrument spec and existing models are updated. Any model and supplier that qualifies the performance specification can serve requests on that instrument. The models listed here were recently active but are not guaranteed at any given time.

3Lab Latest launch prices are single blended rates calculated from the current platform token mix: 10.30% input, 84.07% cache read, 3.39% cache write, and 2.25% output/reasoning. The mix is based on production usage from May 25–June 18, 2026.

Route efficiently to get more out of every dollar.

Point each task to the right instrument to see improvements in performance and cost efficiency.

Triage on Standard,reason on Prime.

Classify requests on Standard first. Send onlythose that need reasoning to Prime.

Retrieve on Standard,synthesize on Prime.

Run retrieval and reranking on Standard. Generatethe final output on Prime.

Prime by default,Max by exception.

Keep your production workload on Prime. Use Maxonly for large inputs or where accuracy is critical.

Learn more on our blog

Read the Research

Commoditizing AI Inference with InstrumentsAre you overpaying for brand-name LLMs?A blend of open-source models matched the top performers