Microsoft has launched “Microsoft-Decision-1,” a decision model that outputs calibrated probabilities rather than text, offered via Microsoft Foundry and soon on OpenRouter. The twist: it is based on Qwen3.5-9B, developed by Alibaba Cloud, and Microsoft says it will soon rebase the model on its own and OpenAI’s models.
The launch is a direct shot at TypeSafe AI’s Jev, which raised $870 million at a $7.5 billion valuation this week on the promise of fast, token-cheap task automation. Microsoft claims Decision-1 is 2.5 times faster than H2O-Lightning-4B and 2.8 times faster than Jev in latency, hits 83.5 percent accuracy across 36 benchmarks, and costs more than 20 times less than OpenAI’s GPT-6 Sol on text classification, with input priced at $0.042 per million tokens and output free.
The field is crowding fast. The Register counts more than 100 decision models now on Hugging Face, with this week’s announcements including OpenAI’s Decisions API in public beta, Cloudflare Clef, and offerings from Liquid AI, Perplexity, and Snowflake. Researchers are studying how errors compound in decision models, a caveat The Register flags as the category races to production.