MICROSOFT AI CHIEF SAYS IT BUILT A BETTER MODEL THAN OPENAI WITH A QUARTER OF THE COMPUTING POWER

What happened: Microsoft AI said it finished pre-training MAI-Base-2, a mixture-of-experts model. CEO Mustafa Suleyman reportedly told staff it beats OpenAI's GPT-5.6 using about a quarter of the computing resources. Key numbers: 80 billion active and 2.8 trillion total parameters. Trained on 16k GB200 chips over 38 days in August and September 2026. Why it matters: The market response was quiet, and the OpenAI comparison rests on private tests, with no release date, pricing, or public benchmarks given. Microsoft says the model is still in post-training, so proof is pending.

Microsoft AI said it finished pre-training MAI-Base-2, a mixture-of-experts model. CEO Mustafa Suleyman reportedly told staff it beats OpenAI's GPT-5.6 using about a quarter of the computing resources.

The market response was quiet, and the OpenAI comparison rests on private tests, with no release date, pricing, or public benchmarks given. Microsoft says the model is still in post-training, so proof is pending.