Microsoft unveils 2T-parameter MAI-Base-2 AI model
Microsoft has introduced MAI-Base-2, a 2-trillion-parameter base model that the company says improves efficiency over leading open-source alternatives. The model, detailed by Microsoft’s AI research group, is not a finished large language model but a foundational prototype designed for developers to fine-tune into full-fledged LLMs.
SiliconANGLE reported the announcement, noting that MAI-Base-2’s efficiency focus contrasts with recent trends in open-source releases. Earlier this month, Reflection AI open-sourced its 501-billion-parameter Beam model, while Mistral introduced its largest AI model to date. Microsoft’s model may reflect a different approach, emphasizing computational practicality.
The timing aligns with Microsoft’s broader AI push, including its streaming voice AI launch earlier this month and its venture arm’s investment in UK event-prediction startup Mantic. While MAI-Base-2’s release lacks the fanfare of open-weight models, its scale and efficiency metrics could influence expectations for base models. Developer adoption may hinge on adaptability and whether Microsoft chooses to share weights or maintain proprietary control.
Sources: siliconangle.com
“Microsoft’s 2T-parameter base model signals a shift toward efficiency over sheer scale, pressuring open-source rivals to optimize rather than expand.”
Read the original reporting
The outlets below did the original reporting.
- Microsoft reveals cutting-edge MAI-Base-2 model with 2T parameters — siliconangle.com
Related briefs
This brief was drafted automatically from the sources above and published under our editorial policy. Spotted an error? Tell us.