Lab notes.
Tutorials on one side, essays on the other, Stoic notes from the shelf, and the weekly Wire on what is moving in AI infrastructure. Written down the week it happened, so it is still true.
Mistral Opens Large 4: A European Trillion-Parameter Model, With One Condition for Hosts
Mistral published the weights of Mistral Large 4, a 1.05T-parameter model trained on 3,800 GPUs in Europe. It closes a month of giant open releases in which licenses, not parameter counts, decide who can serve them.
In an era where market-research PDFs land on desks faster than you can finish a ristretto, prompt-engineered text-summarisation lets analysts trade hours of CTRL-F drudgery for instant, five-sentence clarity on how AI adoption and ramped-up privacy regulation are reshaping customer behaviour.
All tutorials →We had three weeks to build an MVP. No extensions, no excuses. Ambitious? Absolutely. Risky? Of course. But this wasn’t our first rodeo.
Jun 07 · 2 minMistral published the weights of Mistral Large 4, a 1.05T-parameter model trained on 3,800 GPUs in Europe. It closes a month of giant open releases in which licenses, not parameter counts, decide who can serve them.
Issue 044 · Oct 29 · 8 min