Inception Labs releases Mercury 2: first reasoning diffusion language model decoding over 1000 tokens per second on NVIDIA GPUs
Original title:Mercury 2: the first reasoning model fast enough to pick up the phone
AI summary
Inception Labs released Mercury 2, which it calls the world's first reasoning diffusion language model (dLLM), decoding in parallel at over 1000 tokens per second on NVIDIA H100 GPUs, completing a 300-token reasoning chain in about 300ms and fitting within the latency budget of voice conversations.
Key Takeaways
- Inception Labs released Mercury 2,which it calls the world's first reasoning diffusion language…
- decoding in parallel at over 1000 tokens per second on NVIDIA H100 GPUs
- completing a 300-token reasoning chain in about 300ms and fitting within the latency budget of…
Quick Facts
- Published
- Updated
- Source
- Inception Labs:Blog(网页)
- Category
- Model Releases
- AI score
- 67
官方给出扩散架构推理模型在 NVIDIA GPU 上的延迟与定价数据,语音 Agent 开发者可以据此评估是否能替换现有默认模型。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Inception Labs releases Mercury 2: first reasoning diffusion language model decoding over 1000 tokens per second on NVIDIA GPUs. July 14, 2026. https://new.nat6.net/en/news/kfbkxthb845e48ytskg8k3i9f/