Meta's Dual-Track AI: Llama 5 Open, Muse Spark Closed

For years, Meta's AI identity was simple: open weights, for everyone. In 2026, that changed. On April 8, 2026, Meta simultaneously shipped Llama 5 as open weights and Muse Spark as a closed model — splitting its AI strategy for the first time.

The Two Tracks

Llama 5 (open): Sustains the ecosystem breadth and developer goodwill that made Meta a credible AI platform. It keeps Meta central to the open-weight community — even as Chinese models like Qwen and Kimi have overtaken Llama in downloads.

Muse Spark (closed): Pursues the frontier capability Meta needs to keep its Meta AI assistant competitive as ChatGPT and Gemini iterate. This is the pragmatic admission: a purely open strategy wasn't keeping pace at the very top.

Why the Split

The dual-track approach resolves a tension Meta lived with for years. Open weights are great for mindshare and ecosystem — but they also hand your best work to competitors and don't directly monetize. By keeping Muse Spark closed, Meta protects its frontier edge for its own products while Llama 5 keeps the community engaged.

It's Meta trying to have it both ways — and given its resources, it might just pull it off.

The Money Behind It

Funding both tracks at once is staggeringly expensive. Meta's 2026 capex guidance hit $115–135 billion — nearly double the prior year. That's the price of running an open ecosystem and chasing the frontier simultaneously.

Meta Superintelligence Labs

The effort runs through Meta Superintelligence Labs (MSL), the division Zuckerberg assembled in 2025, led by Alexandr Wang (ex-Scale AI, now Meta's Chief AI Officer) as head of the TBD Lab research unit driving the Llama models and Meta AI.

The Bottom Line

Meta's dual-track pivot is one of 2026's most important strategy shifts: open for the ecosystem, closed for the frontier, funded by one of the largest capex budgets in tech history. Whether it can lead both races at once is the open question.

Brandomize helps brands craft AI strategies that balance openness, control, and cost. Open, closed, or hybrid — we help you choose the model approach that fits your goals.