Inception AI Inc. has released its new flagship model, Mercury 3, a diffusion LLM, and says it outperforms competing systems on its own benchmark tests. Across reported comparisons, the company claims Mercury 3 beats OpenAI’s “GPT-6 Luna” and Anthropic’s “Claude Haiku 5.5” on eight of 12 tests and runs substantially faster.
Outlets describe Mercury 3 as a diffusion-based approach that differs from the autoregressive design used by many chatbots, and several focus on speed metrics. One outlet highlights the comparison framing (who it beats and how often), while another emphasizes the technical distinction of diffusion LLMs. A third spotlights a concrete performance figure, citing a throughput rate and context length from Inception’s reported table.
Overall, coverage aligns on the core points attributed to Inception: the model release, the diffusion-LMM framing, the benchmark win count, and claimed speed improvements versus GPT-6 Luna and Claude Haiku 5.5. Differences mainly concern which specific metrics are foregrounded, not whether the comparison exists.
