Inception Launches Mercury 2: Diffusion LLM Hits 1,009 Tokens per Second, 5x Faster Than Speed-Optimized Models

Summary

Inception, a startup founded by Stanford, UCLA, and Cornell researchers behind foundational diffusion model work, launched Mercury 2 on February 24, 2026 — a production-grade reasoning LLM built on a diffusion-based architecture rather than the autoregressive token-by-token generation used by GPT, Claude, and Gemini. Mercury 2 achieves approximately 1,009 tokens per second output throughput, compared with roughly 89 tokens per second for Claude 4.5 Haiku Reasoning and 71 tokens per second for GP