Analysis
Microsoft launched Microsoft-Decision-1 this week, a new "decision model" built on Alibaba's open-weight Qwen3.5-9B, positioning it as a direct, cheaper challenger to Jev, the non-text AI model that Pulse covered hitting a $7.5 billion valuation three weeks after TypeSafe AI's $40 million seed. The Register first reported the launch.
What Changed Since Jev's Valuation Jump
Microsoft's own benchmarks, released alongside the launch, claim Decision-1 runs 2.8 times faster than Jev with 83.5% accuracy across 36 benchmarks and a 92.2% confidence score -- which Microsoft itself describes as second place, meaning Jev still leads on raw accuracy and confidence even as Microsoft undercuts it on speed and price. Decision-1 is priced at $0.042 per million input tokens with free output tokens, and ships first through Microsoft Foundry before reaching OpenRouter. Microsoft VP Achint Srivastava called decision models "quickly emerging as an important new category in AI," pointing to cost efficiency in agentic applications as the selling point.
The bigger shift is how crowded the category has gotten: more than 100 competing decision models are now listed on HuggingFace, barely a month after Jev's launch created the category's first headline valuation. That matters more than any single benchmark -- a $7.5 billion valuation built on a three-week-old model now has to defend its lead against a Big Tech incumbent willing to subsidize pricing with free output tokens, using someone else's open-weight model instead of building its own.
TypeSafe AI has not yet answered with a pricing move of its own, and Jev's accuracy lead has so far only been tested against Microsoft's own benchmark suite, not an independent one.
