OpenAI released three new realtime voice models designed to enable developers to build a new class of voice applications. The models include GPT-Realtime-2 with GPT-5-class reasoning, GPT-Realtime-Whisper for transcription, and GPT-Realtime-Translate for multilingual support.
GPT-Realtime-2 brings advanced reasoning capabilities to voice interactions, matching GPT-5 performance in real-time conversations. The model enables developers to build applications that can understand and respond to complex spoken queries without the traditional latency associated with voice processing.
GPT-Realtime-Whisper handles live speech transcription, converting audio input to text in real time. This component allows for immediate processing of spoken input across applications.
GPT-Realtime-Translate supports translation across 70+ languages, enabling developers to build multilingual voice applications. The feature allows real-time conversation across language barriers without separate processing steps.
OpenAI stated the models will "unlock a new class of voice apps for developers." The release targets the API, making the technology available to third-party developers building commercial and internal applications.
In parallel, OpenAI launched Trusted Contact, an optional safety feature for ChatGPT. The feature allows users over 18 to assign an emergency contact for mental health and safety concerns. The capability expands existing teenage safety options to adult users, enabling designated contacts to be notified when the platform detects potential safety issues.
The Trusted Contact feature is optional and gives users control over their safety settings. When activated, designated contacts can receive alerts if OpenAI's systems detect concerning behavior patterns.
Both announcements reflect OpenAI's focus on expanding voice capabilities while strengthening user safety features. The voice models target developers seeking to integrate advanced audio processing into applications, while the Trusted Contact feature addresses growing concerns about digital safety and mental health support.
AI agents have surpassed human users as the primary consumer of tokens on OpenRouter since early February 2025, with agentic usage jumping 14x while human consumption grew just 2.8x.
A new theoretical study challenges the assumption that AI improves research productivity. Instead of reducing workload, AI could push researchers to launch more projects while quality per publication declines.
Munder Difflin introduces an agent harness platform designed to orchestrate multiple AI agents working in parallel. The tool aims to streamline coordination of autonomous agents for office and business workflows.
A developer spent a week prioritizing OpenAI's Codex over Anthropic's Claude, documenting differences in performance across coding tasks. The experiment garnered significant discussion in the developer community with 113 comments on Hacker News.