Guides, benchmarks & buildlogs
Straight answers on private AI, local models and software engineering, written by the people building it.
Library
Curated guides and benchmarks, kept up to date.RTX 3090 LLM benchmarks: 70B at 48 GB, and what happens with 32 users at once
Local LLM benchmarks on two RTX 3090s, 8B to 70B, measured with CRUCIBLE, our open source tool. 19 to 145 tokens/s for one user, 2,460 tokens/s across 32 at once, plus cited answers over your own documents.
Read article →Local AI server for business: what does it take?
A useful local AI server starts at one 24 GB GPU and 64 GB RAM, from roughly 5,000 euros. Hardware, GPU generations, software, real costs and when the cloud wins.
Read article →Can we use ChatGPT under GDPR?
Yes, often, with the right plan and contracts in place. But not the free version, and not for all data. Here is how to draw the line for your company.
Read article →Feed
Buildlogs and news, newest first.Athlas is now an NVIDIA Inception member.
Athlas has been accepted into NVIDIA Inception, the program for AI startups. What membership gives Blackbox customers, and what it does not change: the models still run on your premises.
Want real help with this?
Athlas builds custom software, AI solutions and digital marketing for ambitious teams. Get in touch and we'll give you a straight answer on whether we can help. You will hear back within 24 hours.