<p>Kyutai has released Voice of Reason, 2 open-weight speech-to-speech models built on GLM-4-Voice-9B. Supervised fine-tuning and reinforcement learning lift spoken GSM8K accuracy from 27.3% to 77.1%. There is no transcription step and no text LLM in the loop. Both checkpoints are on Hugging Face and run on a single H100.</p>
<p>The post <a href="https://www.marktechpost.com/2026/09/22/kyutai-releases-voice-of-reason-a-speech-native-model-that-solves-spoken-math-with-reinforcement-learning/">Kyutai Releases Voice of Reason: A Speech-Native Model that Solves Spoken Math with Reinforcement Learning</a> appeared first on <a href="https://www.marktechpost.com">MarkTechPost</a>.</p>