With Behemoth shelved, the open-weight lead has quietly changed hands
Llama 4 launched in April 2025 and remains the current generation more than a year later. Its two-trillion-parameter Behemoth teacher model, previewed at that launch, has never shipped and is now widely treated as shelved. The open-weight frontier has moved to Alibaba and Mistral by default rather than by contest.
A Western incumbent pausing while a Chinese lab and a European startup sprint is the single biggest reason the open-weight question is genuinely open again. For two years the assumption was that Meta would keep setting the open ceiling and everyone else would work beneath it. That assumption expired quietly, without an announcement.
Behemoth is the specific casualty worth noting. It was the piece that justified the rest of the line — a two-trillion-parameter teacher whose distillations were meant to keep the smaller models competitive for years. Without it, Llama 4's derivatives age against a field that is shipping monthly, and the compounding runs the wrong direction.
What replaces it is less concentrated: Mistral's sparse-MoE Large 3 at 675B total and 41B active with 256K context, the Ministral 3 edge line, Mistral Small 4 from March, and Qwen's steady cadence. No single model dominates, which is healthier for the ecosystem and considerably harder to summarise.
SiliconANGLE — Meta to postpone release of Llama 4 Behemoth model, report claims → · The Daily Star — Alibaba releases Qwen 3.8-Max, its largest AI model yet → · Okoone — Meta puts Llama 4 Behemoth on hold as questions rise →