Meta is adopting a hybrid release strategy for its next-generation AI models under Alexandr Wang, planning to open-source smaller versions while locking down its largest frontier models. Brought in via a $15B Scale AI deal, Wang is designing the new lineup to bounce back from Llama 4's lag, bypassing OpenAI's enterprise turf to push consumer AI directly across WhatsApp and Instagram.
Groq just officially became a supported Inference Provider on the Hugging Face Hub. Developers can now run open-source models like Llama 4 directly through Groq's low-latency LPU chips using HF's web UI or Python and JS SDKs. You can bill through Hugging Face without added markups or plug in your own Groq API key.
A new joint study from Cohere, Stanford, MIT, and Ai2 accuses popular benchmark LM Arena of favoring tech giants like Meta, OpenAI, and Google. The researchers found that big labs were granted extensive private testing to cherry-pick winning variants and received higher battle sampling rates. LM Arena pushed back against the claims as inaccurate, but promised to update its sampling algorithm.
Meta just launched a standalone Meta AI app powered by Llama 4, putting hands-free voice conversations front and center. The app absorbs the Ray-Ban Meta companion controls, includes a native full-duplex voice demo, and taps into your Facebook and Instagram activity for personalized answers. It basically elevates Meta AI from an in-chat bot into a full-blown daily companion.
Google just made Meta’s Llama 4 models generally available as a fully managed API on Vertex AI. Developers can now run Llama 4 Scout and Maverick via standard OpenAI-compatible endpoints without provisioning GPUs or managing infrastructure. Why it matters: it brings Meta's flagship open models straight to enterprise cloud pipelines with zero operational overhead.
LMSYS just dropped over 2,000 battle logs for Llama-4 on Arena and took aim at Meta for transparency issues. The evaluation platform revealed that Meta failed to disclose its experimental Maverick model was heavily tuned to game human preference prompts, prompting an immediate overhaul of leaderboard policies.
Meta’s head of GenAI just denied rumors that Llama 4 Maverick and Scout were trained on test sets to game benchmark scores. The speculation flared over the weekend after users noticed performance gaps across platforms. Al-Dahle blamed the mixed quality on early rollout issues across cloud providers, promising bug fixes over the coming days.
A Meta AI insider just blew the whistle on Llama 4, claiming the model lagged far behind rival open-source models and leadership pushed to blend benchmarks into post-training to fake results. The researcher submitted their resignation in protest and claimed Meta's VP of AI also stepped down for the same reason.
Meta just unveiled Llama 4 Scout and Maverick, marking its official pivot to open-weight, natively multimodal mixture-of-experts architectures. Both models operate on 17B active parameters, with Scout pushing context windows to a massive 10 million tokens, while an upcoming 2T-parameter teacher model, Behemoth, trains in the background.
Meta is reportedly gearing up to launch Llama 4 in just a few weeks, putting a major focus on native voice capabilities. Meta CPO Chris Cox confirmed it will be an "omni" model handling speech and text natively, allowing users to interrupt mid-sentence like ChatGPT's Voice Mode. The rush comes as DeepSeek's open-source gains pushed Meta into overdrive.
Mark Zuckerberg used YouTube's fight against piracy to justify Meta's use of copyrighted datasets during a deposition in an ongoing AI copyright lawsuit. Newly unsealed court filings allege Meta trained its Llama models on pirated e-book repositories like LibGen and Z-Library despite internal legal warnings.
