Microsoft releases Microsoft-Decision-1 built on Qwen3.5-9B

Microsoft has released Microsoft-Decision-1, a decision-scoring model that returns a calibrated probability for each fixed answer option instead of generated text. Post-trained from Alibaba's Qwen3.5-9B, it is available in Microsoft Foundry and OpenRouter and leads Microsoft's 36-benchmark comparison at 83.5% accuracy with 85 ms p50 latency. Pricing is $0.042 per million input tokens with free output.

• 83.5% average accuracy across 36 benchmarks, 85 ms p50 latency
• Post-trained from Qwen3.5-9B with a 32,768-token context window
• $0.042 per million input tokens, output tokens free
• Closed weights; all benchmarks are vendor-run