Inception
AI company building Mercury, a family of fast diffusion-based large language models
About
Inception is an AI research and product company building diffusion-based large language models (dLLMs) for production applications. Instead of generating one token at a time, its Mercury models refine whole outputs in parallel over a few steps, which the company says makes them much faster and cheaper than autoregressive LLMs. Its team includes researchers and engineers from Stanford, UCLA, Cornell, Google DeepMind, Meta AI, Microsoft AI and OpenAI, and its founders' prior work includes diffusion models, Flash Attention, Direct Preference Optimization and Decision Transformers. In October 2026 it released Mercury 3, which it describes as the most capable diffusion LLM to date, running at over 1,000 tokens per second with a 1 million token context window.
- Website
- inceptionlabs.ai
- Added via
- web
- Ownership
- unclaimed
Where to go
For people
- Homepageinceptionlabs.ai
For agents
Nothing listed yet.
Products
- Inception Mercury — Family of commercial diffusion LLMs, including Mercury 3, built for low-latency apps and agents · Api
Social profiles
Links
- Introducing Mercury 3other
- Docsother