Empero unveils Manta long-context model claiming 100M tokens in one prompt, opens early access

HankYeomans · x · 2026-11-07

Startup Empero unveiled Manta, a long-context model claiming to read 100 million tokens in a single prompt — "everything you give it, it keeps", pitched as effectively infinite context.

Key points:

Early access opens to researchers, engineers and teams with concrete long-context use cases via a waitlist, ahead of public release;
Signup review uses an AI model to read applicants' public pages (GitHub, papers, company sites) and write a summary, but a human makes every decision;
Early users may discuss quality but cannot share outputs, prompts, screenshots or numbers until cleared;
The poster quipped: "forget your measly 1M context."

Original post →

Empero unveils Manta long-context model claiming 100M tokens in one prompt, opens early access

More from Models