Post

We've developed a new inference engine, called 'e'. It's the closest thing to alien technology which we have, and pushes 10,500 tok/s aggregate on a single DGX Station with the new @NVIDIAAI Nemotron 3.5 Super VL model. actual.inc/company/blog...

actual.incActual Computer › Case Study : GB300 and Nemotron 3.5 Super VLDepartment-scale AI from one NVIDIA DGX Station: Nemotron 3.5 Super VL at 8,000+ tokens per second, shared by your whole team through api.actual.inc.

16/10/26, 1:59 pm8 reposts122 likes

Tom

@moll.dev

wow okay that's insane

Mrs. Reach / Mrs. Curio

@reachartwork.bsky.social

neat!

Mrs. Reach / Mrs. Curio

@reachartwork.bsky.social

if i have questions about what sort of services actual.inc can run, what's the best way to get them answered? should i dm you, send an email somewhere, or what?

send me a DM!

With Actual's new engine e, one station serves NVIDIA Nemotron 3.5 Super VL at north of 8,000 tokens per second across 128 streams, and tops 10,000 at 256.

∿

@somewhere.systems

An agent swarm needs more than a fast first request. It needs the whole machine to stay responsive as agents join in. In a single stream, e serves Nemotron 3.5 Super VL at 455 tokens per second. A fresh 8K-token prompt gets its first token back in 0.58 seconds.

∿

@somewhere.systems

At 8,000+ tokens per second, one station produces about 691 million output tokens a day. Bought at GPT-6 Luna's list price of $0.50 per million output tokens, that's about $126,000 a year. Flat out at its full 1.6 kW rating, the station uses about $163 a month in electricity.

∿

@somewhere.systems

Today most teams rent their agent swarms from hosted models and pay for every token. At list prices, 32 Luna subagents running around the clock cost $4,686 a month. With Actual, the swarm runs on Nemotron from your DGX Station instead.