Scaled Cognition, which builds models for customer service agents, introduced Voice, a text-to-speech model. The company says it pronounces account numbers and addresses more accurately and costs less than rivals; it is available via API or self-hosted.
OVHcloud's boss revealed that the French cloud provider is developing its own large language model. A first version could launch at the end of November.
ElevenLabs unveiled two speech-to-speech models, Mini and Max, designed as complete voice agents. They are available now and compatible with OpenAI's Realtime API.
Interconnects' regular roundup of open models covers DeepSeek V4.1 Flash 1017, new American-made models and fresh regulatory controls in the US and China. The author frames it as Western open models striking back.
Latent Space's daily roundup notes the releases of DeepSeek V4.1 Flash 1017, MiniMax M3.1 Flash and Mercury 3. It expects none of them to see very wide use right away.
Inception has released Mercury 3, a diffusion-based language model it says generates 1,068 tokens per second with a 1-million-token context window. In Inception's own comparison, it beats GPT-6 Luna and Claude Haiku 5.5 on 8 of 12 benchmarks.