Perplexity is building a very focused inference team that wants to host custom versions of open source models like DeepSeek and Llama with as high throughput and as low latency as possible. This blog post talks about our work on building in house networking. We will be sharing more soon on our post training work.

Great to meet you once again, Prime Minister @narendramodi ji, at Paris!

Is this real ? @realDonaldTrump

@ratkobatinic Okay.

@Klotzkette When it happens, it will happen with a bang.

Careers page for those who want to join and help:

I’m in Paris for a special announcement with Bouygues Telekom CEO Benoît Torloting tomorrow 🇫🇷

@TatTvamAsiOnX There will be more such contests in future

@kaatrinisai Just wait for a bit

@caviterginsoy Weird. Could you send a screen record? Are you on the latest version of the app? @fayfers

@Klotzkette Sure. But it will also use a ton of other sources and can build way more cards than Wikipedia can.

@Pulkit_Saraf Will start showing up for celebrities and well known individuals, and then to businesses and events and related entities.

@_PrasannaLahoti Yeah, that’s the plan. Support something similar to Google.

@a_unbox @perplexity_ai @fayfers

kek