Meta AI / Llama
@tabtab-aiOfficial TabTab account
Track Meta AI / Llama as an AI company/product ecosystem.
Findings5
Meta offers tiered Muse subscription plans and runs Muse on Muse Spark model family
Analysts reporting on Meta’s Muse launch say the company is offering a tiered consumer AI subscription model (free, Powerful at $20/month, Premium at $100/month) and that Muse runs on the Muse Spark model family, including a recent Muse Spark 1.3 update claimed to improve coding and agentic task performance. Wall Street notes the new paid tiering and model updates as a potential revenue channel beyond advertising.
via TabTab1d agoSource
Meta launches Muse AI agent in the US with WhatsApp integration
Meta has launched Muse, a personal AI agent available in the US on iOS, Android, web (muse.ai) and inside WhatsApp that can execute multi-step tasks such as booking travel, filling forms and making purchases. The product is positioned as a shift from Q&A chatbots to agentic AI that runs tasks in isolated cloud VMs and integrates with Stripe for commerce flows, making it a new consumer AI distribution and monetization push for Meta.
via TabTab1d agoSource
Meta acquires Stilla.ai to bring context-aware agent tooling into Muse
Meta has acquired Stilla.ai, a Swedish startup that pulls context from tools like Slack and GitHub to keep teams and agents aligned; the deal will bring Stilla’s team and technology into Meta as it expands Muse and agent tooling, according to reporting.
via TabTab1d agoSource
Meta and This Is Digital launch AI Academy Kenya with pitchathon and masterclass
Meta partnered with African accelerator This Is Digital to launch AI Academy Kenya, a program combining practical AI training, an AI pitchathon (applications open through Sept 30) and an eight-week AI Masterclass starting Nov 3; Meta says the program will use Meta’s AI tools to support local developers and startups.
via TabTab1d agoSource
Court filings show Meta staff discussed using copyrighted books and subtitles to train LLaMA
Published court filings reveal internal Meta discussions about using copyrighted content — including books and subtitles — to train the LLaMA family of models, highlighting legal and ethical debates within Meta over training data choices.
via TabTab1d agoSource
