Tag: llama

Meta is eyeing early 2026 for «Avocado» model, and it might be proprietary

The billion dollar crew at the Superintelligence Labs at Meta are readying their first release.
Meta’s next frontier model might not be open source, rumours say. (Picture: generated)
CNBC reports today that Meta is coming closer to a release of their next frontier model — four months after forming their Superintelligence Labs.

The model is rumored to not be in the Llama family of open source models, as The Zuck himself opined in July that «We’ll need to be rigorous about mitigating these risks and careful about what we choose to open source.»

Therefore, the «Avacado» family of models might well be proprietary, also after Meta saw lots of components of its Open Source models getting incorporated in other advanced models.

Little is known about the upcoming model, apart from the fact that it will be «frontier,» and there has apparently been discussions on releasing it this year — but the official plan is that is scheduled for Q1 2026, according to CNBC.

Read more: Full scoop at CNBC.

Meta discussed moving to closed source models

Meta is holding internal discussion on future closed source models.
AI strategy might be on the move at Meta, according to reports. (Picture: Dylan Tweney, CC BY 2.0)
According to a New York Times article out today, the top leadership of the new Superintelligence Labs at Meta have held internal discussions on ditching open source models.

They have reportedly stopped testing on the open source Behemoth model, after it delivered «underwhelming» performance, TechCrunch reports.

«Unchanged» policy
They also caution that discussions are just discussions and don’t reflect official policy from the social media giant.

Continue reading “Meta discussed moving to closed source models”

Meta announces completion of «Superintelligence Labs»

Zuckerberg announces the completion of the «Meta Superintelligence Labs»
Meta’s historical hiring spree might come to a close now they’ve announced their team. (Picture: Carnaval.com Studios, CC BY 2.0)
After a few hectic weeks of agressively poaching talent from their rivals, and an acqiusition or two, Meta reveals the completion of their new AI lab.

This comes hot on the heels of several months or years of what Zuckerberg deemed lackluster performance from the Llama team, before he decided to get some seasoned pros in.

The new team is headed up Alexandr Wang and Nat Friedman from Scale AI and Github, respectively.

They will be joined by an 11-man team that will spearhead Metas future AI efforts, that reads like a who’s who of significant AI efforts over the last couple of years from Anthropic, OpenAI and even Google’s DeepMind:

Continue reading “Meta announces completion of «Superintelligence Labs»”

Keep AI queries private, Mozilla petitions Meta

It's too easy to share Meta AI queries, and most dont understand they are doing it.
Meta AI users don’t understand what they are sharing, leading to embarrassment — or worse. (Picture: Wesley Fryer, CC BY 2.0)
Mozilla has launched a public petition calling on Meta to shut down its new «Discovery» feed in the Meta AI app, arguing that the tool is «invasive» and violates user expectations of privacy.

The new feature — part of Meta AI’s expansion across Facebook, Instagram, WhatsApp, and Messenger — surfaces old queries in the Discovery feed in Meta AI’s new Llama 4 app.

People don’t know what they are sharing
There is a button in the app where you have to give explicit permission for Meta to share your query , but it seems to be getting lost in the rollout across the Meta universe.

Continue reading “Keep AI queries private, Mozilla petitions Meta”

Meta delays flagship Behemoth model due to performance issues

Prescient words from Facebook's developer conference in 2017.
Prescient words from Meta’s developer conference in 2017. (Picture: Anthony Quintano (CC BY 2.0))
The 2 trillion parameter model was teased at the launch of Llama 4 in early April, and was slated for a release at the LlamaCon, Meta’s AI developer conference in late April.

Now it seems following internal dismay at the model’s performance, it was delayed until June, and has just been postponed again until sometime in the fall or later, writes Reuters.

Continue reading “Meta delays flagship Behemoth model due to performance issues”

Meta AIs new Llama 4 app has access to Facebook, Instagram

Meta AIs standalone app has access to your Facebook and Insta, and will let you share prompts with friends.
Meta AIs standalone will also let you share prompts with friends. (Picture: Meta)
Meta’s Llama 4 might just be the largest AI deployment in the world, having rolled out across Facebook, Instagram and WhatsApp. Now it’s getting its own standalone app.

Meta’s blog post touts the potential of linking it with your other Meta accounts so it can get to know you better, and says the app will remember things about you that you explicitly ask it to retain — or information you already shared on Facebook, for example.

This is slightly off what ChatGPT offers outside the EU, where the app will add all previous chats to its «memory.» The memory feature in Meta AI is also not available in the EU, limiting this service to the USA and Canada for now.

Continue reading “Meta AIs new Llama 4 app has access to Facebook, Instagram”

Meta gamed the LMArena AI benchmarks with new model

People are astounded Meta used a non public, unreleased and optimized model on the industry’s most respected benchmark.
People are astounded Meta used a non public, unreleased and optimized model on the industry’s most respected benchmark. (Picture: Meta)
According to several reports, it seems Meta used an unpublished Llama 4 Maverick model created especially to score well on the LMArena benchmark.

Surprisingly good ranking
The largest selling point for their latest Maverick model was how well it did in precisely this benchmark, scoring just above ChatGPT 4o-latest and and slightly below Gemini 2.5 Pro, considered the current cutting edge of AI engineering.

The fact that Llama 4 Maverick got second place in between these, on the most watched leaderboard in AI, raised a lot of eyebrows over the weekend.

Continue reading “Meta gamed the LMArena AI benchmarks with new model”

Meta drops Llama 4 models, with «mixture of experts»

The Llama 4 models take off to a great start with the benchmarks, runs on little hardware, and is really cost efficient.
The Llama 4 models seem tailor made for STEM benchmarks, runs on little hardware, and is very cost efficient. (Picture: LadyDragonflyCC, CC BY 2.0)
The latest models out of Meta, the Llama 4 Maverick and the Scout, aren’t reasoning models, but put in extra work by using a panel of «experts» instead.

The models are rolling out right now on WhatsApp, Messenger and Instagram Direct, where they are free for everyone with a membership.

Will answer «contentious» questions
According to TechCrunch, the new models will answer questions previous models wouldn’t, responding to «debated» questions and «contentious» prompts «without judgement:»

— We’re continuing to make Llama more responsive so that it answers more questions, can respond to a variety of different viewpoints, and doesn’t favor some views over others, a meta spokesperson told the website.

Continue reading “Meta drops Llama 4 models, with «mixture of experts»”