Inception’s Mercury 2 Named to TIME’s Best Inventions of 2026

Inception, the company behind the first commercial diffusion large language models (dLLMs), announced today that Mercury 2 has been named to TIME’s Best Inventions of 2026 list in the Artificial Intelligence category, recognizing a fundamentally new approach to large language model architecture. See the full list here.

AI is being integrated into every industry in the modern economy through coding assistants, voice agents, search systems, copilots, and agentic workflows that take repeated actions inside larger loops. Each step adds delay, and reasoning adds more. Autoregressive models generate one token at a time, so the more a model thinks, the longer users wait and the more each answer costs.

Mercury 2, launched in February 2026, is the first diffusion LLM built to reason. Unlike traditional autoregressive models that generate text one token at a time, it generates many tokens in parallel and refines them over several passes. The process is much faster – Mercury 2 generates over 1,000 tokens per second, more than five times faster than leading speed-optimized autoregressive models, with roughly ten times the throughput of the GPT Mini-class models it matches on intelligence. It ships with tunable reasoning, a 128K context window, native tool use, and schema-aligned JSON output.

“For a long time, making AI models more capable has meant making them slower and more expensive,” said Stefano Ermon, CEO and co-founder of Inception. “We started Inception because we believed there was a better way. Diffusion models let us rethink how AI generates answers, and we’re showing that speed and intelligence don’t have to come at the expense of each other. It’s exciting to see TIME recognize this approach.”

This recognition comes as the resource cost of AI is under scrutiny, with climbing AI infrastructure expenses and data centers straining to keep up with demand. Every token consumes compute, electricity, and cooling water, and autoregressive generation leaves much of a GPUs time spent waiting on memory instead of doing math. Mercury 2 gives GPUs the dense parallel work they were designed for, producing many tokens per step instead of one. The result is more output per GPU cycle and per unit of energy, on widely available NVIDIA GPUs rather than specialized chips.

Since launching Mercury 2 in February, Inception has launched Mercury 2.5, its most capable production model yet. It is a significant step up in quality over Mercury 2, with the same low-latency, low-cost serving profile. The company also introduced Mercury Voice, a dLLM optimized for voice agents, and Mercury Router, which understands user prompts and routes tasks to the best models according to quality, speed, and cost.

Mercury is available through the Inception API, OpenRouter, and Baseten. TIME’s full Best Inventions of 2026 list is available at ti.me/best-inventions.

About Inception

Inception develops diffusion-based large language models (dLLMs) designed for efficient, low-latency AI applications. While traditional autoregressive LLMs generate text sequentially, Inception’s diffusion-based models generate outputs in parallel, enabling faster inference and improved reliability for real-world use cases like search, voice, and coding. Inception’s Mercury models are available via the Inception API, OpenRouter, and Baseten. Based in Redwood City, California, Inception is backed by Menlo Ventures, Mayfield, Innovation Endeavors, M12 (Microsoft’s venture capital fund), Snowflake Ventures, Databricks Ventures, and individual backers including Andrew Ng, Andrej Karpathy, and Eric Schmidt. For more information, visit www.inceptionlabs.ai.

Contact Information

Hannah Lulinski

VSC for Inception

VSC PR

inception@vsc.co