OpenAI Shares GPT-Live-1's Task Completion and Latency Benchmarks
- Source
- x.com
- Date
How does GPT‑Live‑1 perform on the benchmarks that matter most for production voice agents? We focused on task completion, back-and-forth conversation and turn-taking, how quickly the model responds, and tool use. Here are the results:
Context
OpenAI's developer account published a task-completion chart reporting 86.2% first-attempt (Pass@1) success on Tau3 Voice for GPT-Live-1, versus 45.7% for GPT-Realtime-2.1. Its footnote says GPT-Live-1 was run with Astra at medium reasoning effort as the backend, so the result describes the combined voice-and-reasoning system.
A separate chart reports turn-taking latency on Full Duplex Bench v1 of 0.798 seconds for GPT-Live-1 versus 1.41 seconds for GPT-Realtime-2.1; that chart carries no backend footnote. These vendor measurements address different questions: whether the system completes a task and how quickly the conversation resumes. They support better results on the named tests, not a guaranteed success rate or response delay for every deployment.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
postOpenAI’s GPT-Live-1 debuts at #1 on the Artificial Analysis Speech to Speech Index at a score of 81.5 using Astra as itsArtificialAnlys
articleAnnouncing The Artificial Analysis Speech To Speech IndexArtificial Analysis articles
videoOpenAI Ships GPT-Live-1 Full-Duplex Voice Model in the APIOpenAIDevs
Checking sign-in…
Loading comments…