BREAKING
Microsoft swaps in MAI models
Excel and Outlook go in-house
0
MAI models at Build
0
transcribe languages
0
K
MAI-Thinking-1 context
MAI vs rival models
MAI-Thinking-1
reasoning
●
Beats Claude Sonnet 4.6 in blind eval
●
Matches Opus 4.6 on SWE-Bench Pro
Critics say
caution
●
Called surprisingly mediocre
●
Below Claude and Gemini at frontier
0
$/M input tokens
0
$/M output tokens
Cost-cutting bet, results TBD
AI NEWS BLITZ
Microsoft is quietly replacing OpenAI and Anthropic with its own MAI models.