The Allen Institute for AI has launched Olmo-core 3, an open mixture-of-experts training framework. The redesigned stack keeps experts resident on GPUs to boost efficiency. It achieves up to 2.7 times higher throughput compared to standard Fully Sharded Data Parallel approaches at trillion-parameter scale.
Scaling frontier architectures often strains infrastructure budgets and hardware efficiency. High-throughput open frameworks help enterprise teams train massive mixture-of-experts models with significantly reduced compute overhead.
Microsoft has announced Maia 300, a new proprietary processor engineered specifically for artificial intelligence inference workloads. Meanwhile, Meta has introduced Muse Wonder, categorised as its inaugural frontier model. These releases signal continued momentum in dedicated machine learning hardware alongside advanced foundation model development.
Meta has announced two new frontier artificial intelligence models, named Wonder and Wonder Full. The organisation states that these releases offer performance on par with rival frontier systems from OpenAI and Anthropic.
Meta's Muse Wonder Full (max) model has debuted on the SevenLab AI leaderboard, securing fourth place overall. The ranking uses SevenLab's value-adjusted framework derived from benchmark data by ArtificialAnalysis. This milestone establishes the model as a strong new contender within the top tier of enterprise evaluations.
15 minutes. No pitch deck. Just a conversation about what AI can do for your team.
Talk directly with our AI specialists