|
|
|
|
|
|
by ein0p
503 days ago
|
|
|
Because it's roughly on par with OpenAI's o1 which is not something the gigantic and extremely lavishly resourced GenAI org at Meta has managed to even come close to so far. And DeepSeek's non-CoT model, DeepSeek V3 makes LLaMA4 irrelevant. And they've achieved both of these feats at a small fraction of the overall cost. You can bet there's a lot of panic at Meta and elsewhere. There's no quick fix. Truth in engineering always prevails in the end. |
|
|