AssemblyAI has released its first text-to-speech model, AssemblyAI-TTS-1, now available on OpenRouter. Designed for voice agents, it offers streaming with 90ms latency and over 4,000 voices at $15 per million characters. Third-party benchmarks rank it highly in speed, accuracy, and naturalness.
Streaming Inference
流式推理
Technique outputting intermediate results progressively before full computation completes
Curated Intelligence
OpenAI has officially launched GPT-6 globally, introducing 'Intelligent UI' which dynamically generates interactive charts, buttons, and mini-apps instead of plain text. The model also features a 'think-and-respond' mechanism that cuts wait times by 44% and outperforms GPT-5.6 in complex web search tasks. Paid users receive the GPT-6 Sol tier immediately, while free users get access to GPT-6 Luna shortly after.
The vLLM-Omni team detailed engineering solutions for adapting multi-modal speech models like Qwen3-TTS and VoxCPM2, addressing the conflict between latency-bound Talker stages and throughput-bound Code2Wav modules. By decoupling scheduling strategies and optimizing streaming pipelines, the approach enhances serving efficiency for multi-stage autoregressive models, offering a reusable architectural reference for open-source TTS deployment.