Llama 4, every mention

22 scenes, the whole family · ← back to Llama 4

mentionsepisodesper episode tap a year for its mentions

Verbatim, from the transcripts: the passages where Llama 4 comes up

loading…

▶ 4:33 unnamed speaker The Llama Four moment, that could have been the crescendo. 2 times in the scene

▶ 7:26 unnamed speaker Alexander Wang's hiring followed the disappointing release of Meta's previous model called Llama Four. 2 times in the scene
▶ 7:40 unnamed speaker It also delayed the rollout of its biggest model, called Behemoth, which it never ultimately released.
▶ 11:57 unnamed speaker Meta's new family of AI models can reach the same performance as Kimi K-II with only 30% of compute, and only 10% of the compute to reach, uh, LLAMA-FOR Maverick, so a much more, uh, efficient, uh, efficient, uh, computing frontier here.

▶ 15:07 unnamed speaker And in fact, I would, I would probably argue that, uh, LLMs in the chat form where you have a bunch of information with a, you know, knowledge cut off baked into a, uh, into some big model, a LLAMA-FOR, and it's vended through, you know,…

▶ 1:17:19 John Coogan Lama four on the other hand was a little bit behind. 2 times in the scene

▶ 2:33:55 John Coogan And so the, my, my like naive read on this is that if you had had a top tier open AI researcher at Meta during the Lama four build out, that researcher would have said, Hey, don't focus on pre-training scale as much.
▶ 3:03:40 John Coogan Actually fine tune Lama four or whatever.

▶ 33:52 unnamed speaker And so the, my, my like naive read on this is that if you had had a top tier open AI researcher at meta during the Lama four build out,

▶ 2:50:28 John Coogan Uh, and I mean, they clearly need a lot more reinforcement learning training data for, to improve Lama four, because they need a, they need a reasoning model.
▶ 2:54:20 unnamed speaker And this last Lama four release was

▶ 0:07 John Coogan Can you, uh, do a little introduction, but then I want to hear your reaction to the Lama four news, how you're processing and what it means for your business.
▶ 0:38 John Coogan He says, unpopular opinion right now, but Llamas for ten million token window will finally actually end the long context versus rag debate, but not in the way that other guy is thinking. 3 times in the scene
▶ 3:26 John Coogan Um, so, so let's talk about LamaFour. 3 times in the scene
▶ 8:19 John Coogan Uh, do you think that part of the narrative that we're seeing around Lama IV is just pre-training, scaling, hitting a wall, a need for new algorithms, a need for a deeper focus on reasoning, and maybe even whatever comes after that?

▶ 11:57 unnamed speaker We saw Llama four dropped on Saturday. 2 times in the scene

▶ 7:52 unnamed speaker He claims that Lomafor is going to blow him out of the water, and he's still investing in CapEx.
▶ 19:07 unnamed speaker And they did beat Meta, like Llamaphore still in training, they haven't released a reasoning model, and it certainly would be cool if this had just baked itself into-

▶ 51:58 Jordi Hays Uh, is that these models caught Meta completely off guard and that they perform better than the new Lama four models, which are still being trained.
▶ 54:02 John Coogan And that's what he's talking about when he talks about Lama four Lama four. 3 times in the scene

Turn any episode into a week of clips.