Meta has unveiled Muse Spark, its first flagship large language model built under Chief AI Officer Alexandr Wang’s newly formed Superintelligence Labs — with competitive multimodal, reasoning and agentic performance at a fraction of the compute cost of its older Llama 4 mid-size variant. Meta also confirmed $115-135 billion of 2026 AI capex.
The Model
- Muse Spark — Meta’s first Superintelligence-Labs flagship
- Strong on multimodal perception, reasoning, health, agentic tasks
- Far cheaper compute than the older Llama 4 mid-size variant
- Built under Chief AI Officer Alexandr Wang
The Spend
- $115-135bn 2026 AI capex — nearly double last year
- Aggressive push to close the gap with OpenAI + Google
Why It Matters
- Signals Meta back at the frontier after Llama stumbles
- Cost-efficiency is the new competitive battleground
- Wang’s Superintelligence Labs bet starts paying off
Follow Vibes Uncut Media for continuing AI coverage.
Leave a Reply