MAI-Thinking-1

Training compute

7×10²⁴ FLOP

Parameters

962B

Published

Jun 2, 2026

MAI-Thinking-1 is an AI model developed by Microsoft (United States), first published in June 2026. It works in the language domain, on tasks such as language modeling/generation and question answering.

Training it took an estimated 7×10²⁴ FLOP of compute. The model has 962,000,000,000 parameters. It was trained on roughly 33.5T datapoints. Training ran on 8,192 NVIDIA GB200.

Access: API access. Its weights are not openly released. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Microsoft
Country of organization: United States
Domain: Language
Task: Language modeling/generation, Question answering
Training compute: 7×10²⁴ FLOP
Parameters: 962,000,000,000
Dataset size: 33.5T
Training hardware: NVIDIA GB200
Chips used: 8,192
Training power draw: 19.1 MW
Numerical format: BF16
Model accessibility: API access
Open weights: No
Epoch confidence: Confident