# Mercury 3 [Inception Labs] - Matching Luna/Haiku at 1000 tokens/sec

Digest AI · Generative AI & Models · published 2026-10-19T17:22:01Z · updated 2026-10-19T18:01:12Z

Canonical: https://digestai.news/story/mercury-3-inception-labs-matching-luna-haiku-at-1000-tokens-sec

## Summary

The model supports a 1M‑token context window (up from 260K for Mercury 2.5) and is priced at $0.20 per million input and $0.70 per million output tokens, with a launch discount of 50% to $0.10 per million input and $0.35 per million output tokens; cached tokens are 90% off the input price.

Early‑access partners including TinyFish AI, Octen AI, Lorikeet and others report real‑time use cases. TinyFish AI uses Mercury 3 for voice tutoring with feedback returning in about a second. In a benchmark of 50 live Mind2Web navigation tasks, Mercury 3 completed 50% of the tasks, outperforming GPT‑6.1‑Sol while using the same token volume and agent steps. The model is available via the Inception API Platform, OpenRouter, Baseten and Vercel and is OpenAI‑API compatible.

## Key points

- Launch price is $0.10 per million input and $0.35 per million output tokens, 50% off regular rates.
- Early partners report median generation 0.9 s, end‑to‑end 2.2 s, and 50% task completion on 50 Mind2Web tests.

## Why it matters

A diffusion LLM matching or exceeding the speed of leading autoregressive models opens low‑latency AI services for voice, search and support at lower cost.

## Sources

1. [Mercury 3 Inception Labs - Matching Luna/Haiku at 1000 tokens/sec](https://inceptionlabs.ai/blog/introducing-mercury-3) (inceptionlabs.ai, 2026-10-19)
2. [Inception Releases Mercury 3, Its Flagship Diffusion LLM](https://unite.ai/inception-releases-mercury-3-its-flagship-diffusion-llm) (Unite.AI, 2026-10-19)

## Cite

Digest AI, "Mercury 3 [Inception Labs] - Matching Luna/Haiku at 1000 tokens/sec", 19 October 2026, https://digestai.news/story/mercury-3-inception-labs-matching-luna-haiku-at-1000-tokens-sec

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/mercury-3-inception-labs-matching-luna-haiku-at-1000-tokens-sec.json