Microsoft launched Microsoft-Decision-1, a compact artificial intelligence model designed to choose between predefined options. The new model is built on Alibaba’s open-weight Qwen3.5-9B architecture.
Decision-1 operates 35 times faster than GPT-6 Sol and roughly 100 times faster than GPT-5.6 Luna. For categorising player feedback, the model is approximately 200 times cheaper and more than 14 times faster than GPT-6 Sol.
The company priced the model at $0.042 per 1 million input tokens. When requests are reordered or reworded, Decision-1 changes its decision only an average of 1.3% of the time.
Microsoft also announced the Maia 300 custom AI chip and plans to deploy more than 1 million units next year. Racks containing 96 Maia 300 chips consume about 167 kilowatts of power.
The firm intends to start multi-gigawatt deployments of the Maia 300 chip beginning next year. Microsoft expects capital expenditures to exceed $175 billion during this calendar year.
"It's basically delivering more intelligence for every watt, every dollar, and every byte of memory by optimizing across the entire system," said Rani Borkar, president of Azure hardware systems and infrastructure at Microsoft. Brad Smith noted that AI currently faces fewer safety requirements than milk.
"Right now, your washing machine is subject to more safety rules than AI. Every carton of milk that you see in the grocery store is subject to safety rules and inspection," Smith said. Smith called for global safety rules including an "emergency brake" and a legal requirement for companies to produce incident reports if breached by a rogue AI agent.
He identified Britain as a country well-placed to lead regulation because of the AI Security Institute. AI Minister Kanishka Narayan stated Britain would consider regulating advanced models if the current system proved insufficient.
