Meta Superintelligence Labs ships its first model
Meta's Superintelligence Labs released Muse Spark, their first multimodal AI model under Alexandr Wang's leadership, marking Meta's return to competitive AI development. The newsletter also covers HeyGen's Avatar V model for realistic AI avatars and various other AI industry updates.
Summary
Meta's newly formed Superintelligence Labs, led by Alexandr Wang after Meta's $14.3B acquisition of Scale AI, has shipped its debut model called Muse Spark. This multimodal reasoning model handles voice, text, and image inputs, featuring a contemplating mode where multiple agents collaborate on complex problems. While competitive with frontier models like Opus 4.6 and GPT 5.4 on reasoning tasks, it lags in coding and ARC-AGI 2 benchmarks. The model excels particularly in health reasoning, aligning with Meta's 'personal superintelligence' mission. Unlike Meta's open-source Llama family, Muse Spark is proprietary, though the company hopes to open-source future versions without committing to a timeline. The newsletter also highlights HeyGen's Avatar V, which the company claims is the world's most realistic AI avatar model, capable of eliminating identity drift and creating full video avatars from just 15 seconds of phone recording. Additional coverage includes Anthropic's Claude Managed Agents platform for simplified agent development, a tutorial on creating automated ad generators using ElevenLabs Flows, and various industry updates including Elon Musk's amended OpenAI lawsuit, Perplexity's revenue growth to $450M ARR, and other AI company developments. The newsletter concludes with a reader workflow showcasing how a photographer uses ChatGPT to write Photoshop scripts for simulating in-camera multiple exposure techniques.
About this episode
PLUS: Build an automated ad generator with this new tool
Key Insights
- Meta rebuilt its AI stack from scratch under Alexandr Wang's leadership, moving away from open-source with Muse Spark being proprietary unlike their Llama family
- HeyGen claims Avatar V can eliminate identity drift in AI avatars and separates identity from appearance for the first time, allowing outfit and background changes without re-recording
- Anthropic's Managed Agents platform handles all backend complexity that previously took engineering teams months to set up, with early adopters like Rakuten setting up agents across five departments in about a week each
- Perplexity achieved 50% monthly growth reaching $450M in estimated annual recurring revenue, driven by its Computer agentic system and usage-based pricing model
- Meta's Muse Spark performs competitively with frontier models on reasoning but lags in coding tasks, with particular strength in health reasoning as part of their personal superintelligence mission
Topics
Transcript
Good morning, {{ first_name | AI enthusiasts }}. Last summer, Mark Zuckerberg handed Alexandr Wang the keys to a brand new lab with some expensive, freshly poached talent from top rivals. Today, we got the first look at what that money built. Meta's Muse Spark isn't topping every benchmark, but it puts the tech giant squarely back in the game — with the resources, data, and 3B+ daily users to build on from here. Meta Superintelligence Labs ships first model HeyGen’s Avatar V solves AI’s identity drift Build an automated ad generator with this tool Anthropic simplifies the agent-building system 4 new AI tools, community workflows, and more META The Rundown: Meta’s Superintelligence Labs just rolled out Muse Spark, a…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The Rundown AI
Grok 4.6 storms the AI frontier
SpaceXAI's Grok 4.6 has emerged as a competitive frontier AI model, scoring 61 on benchmarks and undercutting rivals on price at 60% less cost, with Grok 4.7 teased for release in 3-4 weeks. The newsletter covers AI trends including practical guidance on using AI effectively, a personalized cancer vaccine startup for dogs built with AI tools, and updates on competing models and AI products.
Anthropic slips an invisible signature into Claude
Anthropic has announced the implementation of invisible watermarks in outputs from its Claude models to enhance transparency, aligning with the EU AI Act. The move has sparked mixed reactions, as stakeholders navigate the implications for open and private AI models.
Meta returns to its open-source roots
Meta is returning to its open-source roots with the launch of the Muse Glimmer model and plans for Muse Spark 1.2, positioning itself as a competitive force against China's AI dominance. OpenAI has also expanded its security program with a new hacking-tuned model, indicating a growing emphasis on cybersecurity in AI deployments.
OpenAI puts the safety brakes on Astra
OpenAI's Astra model, rumored to be GPT-6, is being treated as a potential 'critical' cyber risk following its success in solving significant problems, prompting increased safeguards and a delay in its rollout. This decision reflects broader concerns about AI's evolving capabilities and associated security risks as evidenced by recent breaches in the AI community.
AI designs viruses never seen in nature
Researchers used AI to design 16 novel viruses that infect only bacteria, successfully creating them in the lab to combat drug-resistant infections. While the work demonstrates beneficial applications, experts warn that the same AI tools could be repurposed to create dangerous pathogens, prompting calls for biosafety guardrails.