Meta targets year end Llama 4.X release

Meta plans to roll out its latest Llama AI model by year end while reorganizing teams and facing staff turnover.

Meta targets year end release for Llama 4.X amid talent churn

Meta is racing to ship Llama 4.X by year end through Meta Superintelligence Labs, specifically a subgroup called TBD. The project aims to produce a production ready version of Llama 4.X, which insiders sometimes refer to as Llama 4.5. The move comes after Meta released Llama 4 models in April, a launch that some developers felt fell short on real world tasks like coding and reasoning. The TBD team is also tasked with reviving the earlier Llama 4 effort and fixing bugs.

The push to move quickly signals Meta’s willingness to bend its structure to speed. Reorganizing AI work around four pillars—training, research, product and infrastructure—shows a company trying to align big ambitions with clearer responsibilities. Yet talent remains a bottleneck. Meta has offered multimillion dollar packages to lure researchers from rivals, a strategy that raises questions about long term costs and loyalty. If the next wave of models stalls or falters, the momentum may be hard to regain.

Highlights

Ambition in AI is growing faster than the systems that govern it.