Microsoft AI (MAI): the 3T-parameter in-house pre-training claim

status: watch, opened 2026-10-15, last moved 2026-10-15. AI microsoft models

• The thread: Microsoft AI (MAI) claiming in-house frontier pre-training — state-of-the-art pre-trained base models in the 3T-parameter weight class, with post-training results promised.
• 2026-10-15 (first record): ~12:16 PM ET, Satya Nadella RTing Mustafa Suleyman (https://x.com/satyanadella/status/2110778673850941834): "We now have state-of-the-art pre-trained base models in the 3T-parameter weight class, built entirely in-house at MAI. Post-training results coming soon — stay tuned!" with a technical blog post (https://microsoft.ai/news/frontier-pre-training-at-microsoft-ai/). Debut thread — the record's Microsoft AI coverage sat on silicon (Maia 300) and governance posture, not an in-house frontier-model line. Announced, not shipped: no third-party benchmark or eval artifact in capture; "state-of-the-art" is the lab's own modifier in its own blog. Watch: post-training results and benchmarks, independent eval placement, exposure surface (Copilot/Azure).