vibagor441 (@vibagor44145276) 's Twitter Profile
vibagor441

@vibagor44145276

anonymous llama herder

ID: 1830089731142713344

calendar_today01-09-2024 03:46:14

1,1K Tweet

920 Takipçi

39 Takip Edilen

vibagor441 (@vibagor44145276) 's Twitter Profile Photo

occasionally schmidhuber shows up on my radar and I check out his old papers. does anyone else feel like you're reading a patent application when you read his papers? not sure if bc german writing style, but it feels like it was written to be as broadly applicable as possible.

vibagor441 (@vibagor44145276) 's Twitter Profile Photo

From Meta Marketing: Llama 4 to be released on April 10th. Don't know the final marketing name to be used, but we can expect another one of those 4, 4.1, 4.2, etc... series this year. ** everything subject to change

vibagor441 (@vibagor44145276) 's Twitter Profile Photo

Subject to change is subject to change - now the first models will be coming out on April 3rd to beat the rumored OAI release on April 4th. Reasoners will come out on April 10th.

vibagor441 (@vibagor44145276) 's Twitter Profile Photo

Dust is settling: next week, two models (with reasoner variants). And base models, ofc. 1. 17B active, 109B total, 16 experts, 10M (!!) context length. 2. 17B active, 400B total, 128 experts, 1M context length. The name is weird (maybe a code name?) so I won't say nothing yet..

AI at Meta (@aiatmeta) 's Twitter Profile Photo

Today is the start of a new era of natively multimodal AI innovation. Today, we’re introducing the first Llama 4 models: Llama 4 Scout and Llama 4 Maverick — our most advanced models yet and the best in their class for multimodality. Llama 4 Scout • 17B-active-parameter model

Today is the start of a new era of natively multimodal AI innovation.

Today, we’re introducing the first Llama 4 models: Llama 4 Scout and Llama 4 Maverick —  our most advanced models yet and the best in their class for multimodality.

Llama 4 Scout
• 17B-active-parameter model
vibagor441 (@vibagor44145276) 's Twitter Profile Photo

Reasoners continue to be delayed due to bad benchmark scores. There will be a release of guardrail models (meh) on LlamaCon though.

vibagor441 (@vibagor44145276) 's Twitter Profile Photo

Well, the open source model is in June, but I didn't think it would be this early? (I don't think it's the open source model.)