# RuntimeWire — Inception launches Mercury 3 at 1,068 tokens per second

- Company: RuntimeWire (runtimewire.com)
- Announced: 2026-10-19T17:06:00+00:00
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://runtimewire.com/article/inception-mercury-3-diffusion-llm-launch
- Record: https://forck.live/items/24735-inception-launches-mercury-3-at-1-068-tokens-per-second
- Subject: Reporting
- Models affected: Mercury 3

Inception launched Mercury 3 on October 19th, pitching its diffusion-based language model as a faster alternative to conventional systems that generate text one token at a time. CEO and co-founder Stefano Ermon says it delivers more than 1,000 tokens per second while improving on the company's earlier models. The launch tests whether diffusion can scale into general agent and business workloads while keeping its speed advantage. Its published results and partner examples are company-reported, so adoption will depend on how the model performs in developers' own tasks.

## Evidence

Verbatim from https://runtimewire.com/article/inception-mercury-3-diffusion-llm-launch:

> Inception launched Mercury 3 on October 19th, pitching its diffusion-based language model as a faster alternative to conventional systems that generate text one token at a time. CEO and co-founder Stefano Ermon (@StefanoErmon) says it delivers more than 1,000 tokens per second while improving on the company's earlier models.

---

Record: https://forck.live/items/24735-inception-launches-mercury-3-at-1-068-tokens-per-second
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md