Microsoft introduces Decision-1 for classification and routing
Decision-1 from Microsoft is based on Qwen3.5-9B and, according to the company, is designed for classification, evaluation and routing. It is available through Microsoft Foundry and OpenRouter at 0.042 USD per million input tokens; output tokens are free.
Microsoft introduced Decision-1, a model based on Qwen3.5-9B. According to the company, it is designed for fast, structured decisions: classification, evaluation and routing. The model is available through Microsoft Foundry and OpenRouter. A million input tokens cost 0.042 USD; output tokens are free.
According to Microsoft, Decision-1 achieved the highest accuracy among the tested models across 36 benchmarks covering nearly 150 000 questions. The company also reports that it was 2.5 times faster than H2O-Lightning-4B, which ranked second for speed. The comparison did not include Clef models from Cloudflare. The company describes the ability to control agents in complex environments as a potential capability of the model.
Why it matters
Developers can try a specialized model available through two services for individual decision steps in applications. However, the comparison results from the model provider do not cover all alternatives and do not, on their own, demonstrate accuracy on specific business tasks.
Release card
Decision-1
Microsoft
- Price
- The price is 0.042 USD per million input tokens; output tokens are free.
- Availability
- The model is available through Microsoft Foundry and OpenRouter.
- Srovnání rychlosti 2,5krát vyšší rychlost According to Microsoft, Decision-1 was 2.5 times faster than H2O-Lightning-4B, the second-fastest model in the speed comparison.
- The model is designed for classification.
- The model is designed for evaluation.
- The model is designed for routing.
According to Microsoft, Decision-1 achieved the highest accuracy among the tested models across 36 benchmarks covering nearly 150 000 questions. Clef models from Cloudflare were not included in the comparison.
The card summarizes information from the article and any dated corrections, with a link to the original source. It is not our assessment of the model. It does not yet have a dedicated editorial profile. Model selection and other announcements →
Relevant practical impact
What this means
For a business
For business processes involving large numbers of classification or routing decisions, token costs can be planned based on input volume, because output tokens are not charged under the stated pricing.
ProcessesCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.