Introducing GPT‑Live

We’re launching GPT‑Live, a new generation of voice models that make talking with AI feel much more like having a real conversation.

GPT‑Live is built on a full-duplex architecture, meaning it can listen and speak at the same time. During conversations, GPT‑Live can show it’s paying attention with phrases like “mhmm” or “yeah”, engage in quick back-and-forth, or just stay quiet when you need a moment to think. The result is a voice experience that is refreshingly easy to talk to.

GPT‑Live is also our smartest voice model yet. For questions that require web search, deeper reasoning, or more complex work, it delegates to our latest frontier model behind the scenes and brings the result back into the conversation when it’s ready. While it works, GPT‑Live can keep talking with you and maintain the flow of conversation. At launch, GPT‑Live will use GPT‑5.5 in the background. As we release new frontier models, we’ll continuously update the model used by GPT‑Live.

These advances power a new ChatGPT Voice experience that is more intelligent and natural to use. Over time, we believe this research will also unlock the ability to use voice for increasingly complex, longer-running, and more agentic work.

We’re beginning to roll out two versions of GPT‑Live – GPT‑Live‑1 and GPT‑Live‑1 mini – to ChatGPT users globally today. We also plan to bring them to the API soon, and developers and enterprises can sign up to be notified using this form⁠.

Our vision is to enable truly natural human–AI interaction: a world where collaborating with AI feels as fluid and responsive as working with another person, while reasoning and complex task execution happen seamlessly in the background.

Older generations of voice AI systems brought us closer to that vision, but with important tradeoffs.

Cascaded voice systems rely on a series of models acting one after another to process each turn. The original ChatGPT Voice chained three models together: a speech-to-text model to transcribe your speech, a large language model to produce a response, and a text-to-speech model to convert it back into speech. This approach enabled us to talk to frontier AI models for the first time, but the complexity came at a cost: information could be lost across models, and responses were slow and stilted.

Turn-based voice models like ChatGPT Advanced Voice Mode processed and generated audio within a single model, reducing latency and making conversations smoother — but they still operated through discrete turns. The model had to wait for the user to stop speaking before responding, resulting in rigid back-and-forth. In addition, because turn detection is based on silence, even a brief pause or background noise could be mistaken for the end of turn — causing the model to interrupt at unnatural times.

GPT‑Live addresses these limitations through two architectural changes.

First, we built GPT‑Live for continuous interaction using a full-duplex architecture. Instead of processing a sequence of separate messages, GPT‑Live continuously processes input while generating output. The model can therefore make interaction decisions many times per second: whether to speak, continue listening, pause, interrupt, or invoke a tool.

This allows the model to engage in more natural back-and-forth, maintain a better sense of time, and even perform live translation.

Second, we decoupled GPT‑Live — which handles continuous interaction — from deeper work. When a question requires search, reasoning, or more agentic capabilities, GPT‑Live can delegate the task to another model like GPT‑5.5. This allows it to keep the conversation going, even as it handles multiple tasks in the background.

This architectural change also allows GPT‑Live to continuously use the latest models and agents, combining frontier intelligence with natural interaction.

We built new human evaluations to measure pleasantness and the flow of conversation. In these head-to-head comparisons, GPT‑Live‑1 and GPT‑Live‑1 mini are strongly preferred over Advanced Voice Mode in matched 5–10 minute conversations that measure overall preference, turn-taking, interruptions, conversational flow, and how natural each interaction felt.

• GPQA: GPT‑Live‑1 substantially outperforms Advanced Voice Mode on GPQA, which tests expert-level scientific reasoning across biology, chemistry, and physics.
• BrowseComp: GPT‑Live‑1 shows strong gains over Advanced Voice Mode on BrowseComp, which tests agentic web search and the ability to find difficult-to-locate information.
• τ³-Voice Telecom (internal variant)**: GPT‑Live‑1 outperforms Advanced Voice Mode on τ³-Voice Telecom, which tests voice agents on realistic, multi-turn telecom support tasks.

Saben minggu, luwih saka 150 yuta wong ngobrol karo ChatGPT nganggo fitur kaya Voice lan Dictation. Dheweke nggunakake kanggo njaluk pitulungan saben dina tanpa nyekel piranti, latihan basa, nyritakake dongeng sadurunge turu, utawa mung ngobrol nalika mangkat-pulang kerja.

Wiwit dina iki, nalika sampeyan nutul tombol Voice kanggo ngobrol karo ChatGPT, sampeyan bakal oleh pengalaman sing luwih apik didhukung GPT‑Live—kanthi pacelathon luwih alami, wangsulan luwih pinter, pangrungon luwih apik, lan respons visual.

Ngobrol karo ChatGPT saiki kudu krasa luwih kaya pacelathon tenan. Sampeyan bisa nyela nganggo pitakonan, mandheg sedhela kanggo mikir, utawa njaluk ChatGPT ngomong luwih alon. ChatGPT bakal nanggapi kanthi alami marang omongan sampeyan nganggo frasa kaya “mhmm” utawa “ngerti,” supaya sampeyan ngerti yen dheweke tetep nyimak. Kita uga wis ngremaster sangang swara khas ing ChatGPT kanggo GPT‑Live.

ChatGPT Voice saiki bisa nggunakake model tercanggih paling anyar saka kita, supaya sampeyan entuk wangsulan luwih pinter nalika dibutuhake. Sampeyan uga bisa milih tingkat nalar sing cocog karo kabutuhan: Instant kanggo respons cepet, utawa Medium lan High nalika sampeyan pengin ChatGPT mikir luwih suwe.

Yen sampeyan butuh wektu kanggo mikir, ChatGPT Voice saiki bakal ngenteni tinimbang langsung mlebu lan nyela. Yen sampeyan njaluk supaya meneng lan ngrungokake, dheweke bakal nindakake. Lan nalika ana rame ing latar, kaya lalu lintas sing liwat utawa obrolan ing cedhak, ChatGPT luwih bisa fokus marang swara sampeyan tinimbang keganggu.

Sawetara wangsulan luwih migunani yen bisa dideleng. Nalika sampeyan lagi ngobrol, ChatGPT saiki bisa nampilake kertu visual sugih kanggo topik kaya cuaca, saham, olahraga, lan liya-liyane. Voice uga tetep ndhukung panelusuran, memori, gambar, lan unggahan file.

Asile yaiku pengalaman ChatGPT Voice sing krasa luwih alami, luwih kuwasa, lan luwih migunani ing urip saben dina.

GPT‑Live dirancang supaya aman kanthi standar. Iki dibangun saka kemajuan keamanan ing model paling anyar saka kita, ditambah latihan keamanan khusus ing area risiko utama lan pengaman anyar sing dirancang khusus kanggo swara.

Supaya luwih nggambarake cara wong nggunakake swara ing kahanan nyata, kita miwiti kanthi nggedhekake pangujian keamanan kanggo nyakup evaluasi anyar sing asli audio. Kita uga nggawe evaluasi sintetis sing nggunakake audio sing digawé kanggo luwih fokus kanthi intensif ing area keamanan utama, adhedhasar pelajaran saka mode swara canggih. Area kasebut kalebu cilaka marang diri, psikosis lan mania, ketergantungan emosional marang AI, kekerasan, lan konten seksual. Para ahli internal uga nindakake red-team marang model kanggo risiko sing unik ing swara.

Ing pangujian kita, GPT‑Live kinerjane sebanding utawa luwih apik tinimbang mode swara canggih ing meh kabeh area sing dievaluasi. Sampeyan bisa maca luwih lengkap babagan pangujian lan pengaman kita ing kertu sistem⁠(mbukak ing jendhela anyar) GPT‑Live.

Amarga obrolan swara lumaku wektu nyata, kita uga mbangun pengaman sing bisa tumindak nalika model lagi ngomong. Nalika sistem ndeteksi output sing bisa ora aman, sistem bisa ngarahake model menyang respons sing luwih aman, nampilake pesen utawa sumber daya keamanan tambahan, utawa mungkasi obrolan swara ing kasus sing risikone luwih dhuwur. Kanggo pacelathon sing nyangkut cilaka marang diri, kita ngadaptasi alur dhukungan ChatGPT kanggo swara, kalebu nawakake dhukungan hotline krisis sing wis ditinjau ahli.

Kita ngrancang perlindungan tambahan kanggo ndhukung pangguna remaja, lan nglatih perilaku sing cocog umur langsung ing model kanggo nyuda risiko respons sing ora pantes. Wong tuwa bisa milih apa remajane bisa nggunakake ChatGPT Voice liwat kontrol wong tuwa, lan wong tuwa sing disambungake bisa diwenehi kabar ing kahanan risiko luwih dhuwur sing ana tandha potensi cilaka marang diri utawa niat bunuh diri.

Kita uga lagi ngluncurake pangukuran jangka luwih dawa lan pemantauan sawisé peluncuran sing fokus marang ketergantungan emosional supaya terus nambah pangerten lan nyempurnakake pengaman. Kanthi nerusake riset sadurunge babagan panggunaan afektif lan kesejahteraan emosional, iki bakal mbantu kita ngenali pola anyar lan ningkatake cara sistem nanggapi interaksi sing sensitif sacara emosional.

Pungkasan, GPT‑Live dirancang kanggo pacelathon, dudu kanggo niru swara. GPT‑Live nggunakake sakumpulan swara sing wis ditemtokake ing ChatGPT, kanthi pengaman kanggo nyegah niru swarane wong nyata.

Kita setya ndhukung keamanan lan kesejahteraan, lan bakal terus nguwatake perlindungan iki nalika sinau saka panggunaan nyata.

GPT‑Live saiki lagi diluncurake kanggo pangguna ChatGPT ing saindenging jagad liwat iOS, Android, lan ChatGPT.com⁠(mbukak ing jendhela anyar). GPT‑Live‑1 bakal dadi model standar sing nguwasani ChatGPT Voice kanggo pangguna Go, Plus, lan Pro, lan GPT‑Live‑1 mini bakal dadi standar kanggo pangguna Free. Rincian kasedhiyan liyane bisa ditemokake ing Pusat Bantuan⁠(mbukak ing jendhela anyar) kita.

Kita wis ngoptimalake GPT‑Live kanggo sawetara basa sing paling populer ing ChatGPT. Kanggo basa tartamtu, model bisa duwe aksen non-natif utawa kekurangan ing kelancaran. Kita aktif ngupaya ningkatake pengalaman ing macem-macem basa.

Nalika diluncurake, GPT‑Live durung ndhukung swara nganggo video utawa berbagi layar ing ChatGPT, nanging kita lagi ngupaya ngenalake kemampuan iki kanthi cepet. Sampeyan isih bisa ngakses versi lawas ChatGPT Voice, kalebu Standard lan mode swara canggih, ing panggonan fitur iki kasedhiya.