Voice of Reason: reasoning over spoken math
In a September 21, 2026 post, Kyutai Labs says its model solves spoken math problems without an intermediate transcript and scores 77.1% on GSM8K.
Source: Voice of Reason responde a problemas matemáticos falados em voz (x.com). Text prepared with AI from this source.
What happened and what to do
In a post dated September 21, 2026, Kyutai Labs describes Voice of Reason, a model that receives spoken math problems and answers by voice without an intermediate transcript or a text language model in the process. The post reports a 77.1% score on GSM8K, compared with 27.3% for GLM-4-Voice. It presents the approach as speech-native, reasoning directly from audio input.
A company assessing voice support or spoken interfaces could build a prototype for bounded tasks, log inputs and responses, and measure accuracy, latency, and failure cases in its own scenarios. An evaluation and monitoring pipeline can help determine whether direct speech processing fits the use case, without assuming GSM8K results will carry over to other applications.
How the consultancy can help
Wendelmaques can diagnose the opportunity and its data, security, and operational requirements, define an evaluation for the company's case, and propose a scoped implementation with monitoring and operation.
Next step
Send a short description of your voice use case to receive a proposal scoped for diagnosis and implementation.
Let's assess your project
I reply with scope and a proposal.