RT @jrysana: gm

RT @jrysana: Average end to end latency on our benchmarks now below 33ms, well over 10,000 t/s throughput and magnitudes faster than any ot…

RT @jrysana: new peak on small checkpoint

RT @jrysana: Inversion grammar compiler ~10x faster now, around 10,000,000x faster than alternatives on heavy workloads.

RT @jrysana: we made AI structured outputs 1 MILLION times faster

RT @jrysana: here’s a look at how you’ll configure Ana, your new personal assistant built on the fastest AI in the world @RysanaAI Inversio…

RT @anarysana: a little bit about me

@anarysana @calvinducharme @jrysana it's true.

RT @jrysana: say hi to Ana - your new personal assistant, with the fastest general AI in the world powered by @RysanaAI Inversion. she's e…

RT @jrysana: don't blink

ix xxiv

RT @jrysana: Introducing Inversion, our family of structured LLMs. Our first generation models excel in structured tasks, offering unmatch…

@thomaschubak yes

@knowrohit07 we noticed other companies claiming to do large amounts of "tokens per second" vaguely, and it seems like they were really talking about "across many separate user requests" so wanted to clarify

@_Mira___Mira_ try 10 microseconds

@tawnniee @sama ours is ~1,000,000x faster

Function calling at 60 FPS (<15ms end to end) Game on 🔫

@draseac @Deor By shifting from AI calling 1 of N predefined tools to being able to compose a set of atomic tools together in any valid program just like a human programmer, the space of unique problems that can be solved will grow larger than ever.

@mvlcfr190 This message is from a user (not us) calling our global API from across countries.