Intel / article

Mercury 2 The First Reasoning Model Fast Enough To Pick Up The Phone

Source: inceptionlabs.ai
Author: Inception
Date: Published Jul 14, 2026

Context

Inception's Mercury 2 account explains how its Mercury 2 , which generates text in parallel, could make reasoning more usable in voice agents. The model repeatedly refines a noisy sequence toward clean text, working on multiple positions in each pass instead of producing every sequentially.

The company reports generation above 1,000 tokens per second and uses a 300-token reasoning trace to illustrate roughly 300 milliseconds of generation time. That calculation concerns the language-model portion of a call. A complete voice system also transcribes incoming speech and synthesizes the response, so this figure alone does not establish the delay a caller experiences.

Terms in this piece · Glossary

LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.

Read the source www.inceptionlabs.ai