7The Batch (DeepLearning.AI)·
OpenAI releases GPT-Live-1 voice models with full-duplex audio and background reasoning via GPT-5.5
OpenAI released GPT-Live-1 and GPT-Live-1 mini on July 8, 2026, replacing Advanced Voice Mode with a full-duplex voice system that processes audio input and output simultaneously. When deeper reasoning is needed, the voice model delegates to GPT-5.5 or GPT-5.5 Thinking in the background while continuing to speak. GPT-Live-1 at high reasoning scored 84.2% on GPQA versus 45.3% for its predecessor AVM, and human raters preferred it 75.7% of the time. The release also covers Andrew Ng's editorial on AI's labor market effects and a segment on detecting manipulative model behavior.
Related guides (4)
Related events (8)
OpenAI GPT-Live Pairs Full-Duplex Voice Models with GPT-5.5 Reasoning Backend
OpenAI released GPT-Live-1 and GPT-Live-1 mini on July 8, 2026, replacing Advanced Voice Mode with a full-duplex voice system that processes audio continuously and delegates harder queries to GPT-5.5 in the background. The architecture separates a real-time conversational voice model from a reasoning model, with user-selectable reasoning effort levels (Instant, Medium, High) routing to GPT-5.5 Instant or GPT-5.5 Thinking accordingly. Performance gains are substantial: GPQA scores jumped from 45.3% (AVM) to 84.2% (GPT-Live-1 at high reasoning), and BrowseComp improved from 0.7% to 75.2%. The system is live globally on iOS, Android, and ChatGPT.com for paid plans, though no developer API has shipped yet.
OpenAI releases GPT-Realtime-2.1 and GPT-Realtime-2.1 mini for voice applications
OpenAI released GPT-Realtime-2.1, an updated realtime reasoning model with improvements to alphanumeric recognition, silence and noise handling, and interruption behavior. A companion model, GPT-Realtime-2.1 mini, was also released as a faster, lower-cost distilled variant for realtime voice use cases. The releases represent incremental improvements to OpenAI's realtime voice API tier rather than a flagship capability shift.
OpenAI introduces GPT-Live, a new generation of voice models for ChatGPT
OpenAI announced GPT-Live, a new generation of voice models designed for natural human-AI interaction, now powering ChatGPT Voice. The announcement comes from OpenAI's official blog, indicating a production-grade voice capability release. This represents a significant update to OpenAI's real-time voice interaction stack.
OpenAI launches GPT-Live-1 as new ChatGPT Voice engine with simultaneous listen/speak capability
OpenAI is rolling out GPT-Live-1 to power ChatGPT Voice for paid users, with GPT-Live-1 mini for Free users. Both models support simultaneous listening and speaking, enabling more natural turn-taking and interruptions. GPT-Live-1 integrates with web search, memory, visual widgets, and multimodal (text and image) input within a single chat session, though video and screen sharing remain exclusive to the existing Advanced Voice Mode. The rollout targets consumer plans on chatgpt.com and mobile apps, excluding Business, Enterprise, and Edu workspaces at launch.
OpenAI Updates Audio Models That Reason, Transcribe, and Translate
OpenAI introduced three new audio models in its Realtime API: GPT-Realtime-2 (speech-to-speech with five configurable reasoning effort levels), GPT-Realtime-Translate (70+ input languages), and GPT-Realtime-Whisper (transcription). GPT-Realtime-2 operates as an end-to-end audio model including reasoning, with latency ranging from 1.12 seconds at minimal effort to 2.33 seconds at high effort. Benchmark results are mixed: it leads Scale AI's Audio MultiChallenge and Artificial Analysis Conversational Dynamics but trails Step-Audio R1.1 Realtime and Grok Voice Think Fast 1.0 on speech reasoning and agentic tasks. The configurable reasoning-latency tradeoff is positioned as a key differentiator for voice agent applications.
OpenAI releases GPT-5.4 with integrated reasoning, coding, and agentic capabilities in ChatGPT
OpenAI released GPT-5.4, a frontier model that consolidates recent advances in reasoning, coding, and agentic workflows, incorporating the coding capabilities of GPT-5.3-Codex. The ChatGPT deployment introduces a 'Thinking' mode that surfaces an upfront reasoning plan mid-response, allowing users to redirect the model before it completes. Additional improvements include enhanced deep web research for specific queries, better long-context retention during extended thinking, and improved context window management for faster, higher-quality outputs.
OpenAI adds GPT-5.5, reasoning controls, and speech output to ChatGPT workspace agents
OpenAI is rolling out several capability upgrades to workspace agents in ChatGPT Business, including support for GPT-5.5 with configurable reasoning effort, guided agent setup flows, speech/audio output, and smarter Slack thread reply logic. Creators can now tune reasoning intensity per agent and opt into or out of thread-following behavior in Slack. The update expands the enterprise agentic surface of ChatGPT with both model-tier access and new modality support.
OpenAI releases GPT-5.5 and GPT-5.5 Pro to the API with 1M token context and built-in agentic tools
OpenAI released GPT-5.5 and GPT-5.5 Pro to the Chat Completions and Responses API, positioning them as frontier models for complex professional work and compute-intensive tasks respectively. GPT-5.5 supports a 1M token context window, image input, structured outputs, function calling, built-in computer use, hosted shell, MCP, web search, and Skills. Notable behavioral changes include reasoning effort defaulting to medium and extended-only prompt caching support.