Meta promised developers access to its next major AI model months ago. The Wall Street Journal reports the company has delayed the release multiple times now, with no new date announced.
These delays hit the open-source AI market directly. Meta's Llama model family has become infrastructure for thousands of teams - startups, enterprise engineering groups, and researchers who want to run AI on their own servers, customize models for specific tasks, or avoid paying per-query API fees. Every month of delay is another month that alternatives like Mistral, Qwen, and DeepSeek get more embedded in products that might have otherwise run on the next Llama release.
The Journal frames this as a recurring pattern rather than a one-time slip - "keeps delaying" is the language used - which points to genuine engineering challenges rather than a deliberate strategic hold. Meta hasn't offered a public timeline or an explanation for what's causing the holdup.
Developers with deadlines should treat this as a planning signal. Meta's existing open models are still available, but teams waiting specifically on the next generation are now operating on an indefinite timeline - and the teams they're competing with aren't waiting.