Mercury: Ultra-Fast Language Models Based on Diffusion

TL;DR

Mercury introduces diffusion-based large language models optimized for coding, achieving unprecedented speed and competitive quality, with practical deployment and real-world validation on developer tools.

Contribution

This work pioneers diffusion-based LLMs for coding, demonstrating significant speed improvements and competitive performance compared to existing models.

Findings

Mercury Coder Mini and Small achieve 1109 and 737 tokens/sec respectively.

Models outperform speed-optimized frontier models by up to 10x.

Mercury models rank second in quality and are the fastest overall on Copilot Arena.

Abstract

We present Mercury, a new generation of commercial-scale large language models (LLMs) based on diffusion. These models are parameterized via the Transformer architecture and trained to predict multiple tokens in parallel. In this report, we detail Mercury Coder, our first set of diffusion LLMs designed for coding applications. Currently, Mercury Coder comes in two sizes: Mini and Small. These models set a new state-of-the-art on the speed-quality frontier. Based on independent evaluations conducted by Artificial Analysis, Mercury Coder Mini and Mercury Coder Small achieve state-of-the-art throughputs of 1109 tokens/sec and 737 tokens/sec, respectively, on NVIDIA H100 GPUs and outperform speed-optimized frontier models by up to 10x on average while maintaining comparable quality. We discuss additional results on a variety of code benchmarks spanning multiple languages and use-cases as well as real-world validation by developers on Copilot Arena, where the model currently ranks second on quality and is the fastest model overall. We also release a public API at https://platform.inceptionlabs.ai/ and free playground at https://chat.inceptionlabs.ai