Unverified50% confidenceReleaseExact time
Mistral released Voxtral TTS in March of this year, using a decoder-only transformer backbone that autoregressively predicts continuous latent variables, which are then reconstructed by a neural codec.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/3/2026
First Seen
Valid until: 1/1/2027
Sources
Related Entities
Related Claims
UnverifiedThe Mistral series of models is characterized by being open-source and efficient, suitable for local deployment and domain fine-tuning on mathematical corpora; the 'stral' in the name Leanstral may originate from this.62% similarUnverifiedMistral, launched by a French company founded by former DeepMind and Meta researchers, has a Mixtral series that uses a Sparse Mixture of Experts (Sparse MoE) architecture, activating only a subset of parameters during inference.61% similarUnverifiedGLM (the model series launched by Zhipu AI) has become available on Mistral's platform.60% similarUnverifiedMistral、Zephyr等多个开源模型采用DPO作为对齐方案59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/969743API
curl https://kongchang.com/api/v1/knowledge/claims/969743MCP
get_claim(id=969743)