OpenAI's GPT-Live-1 debuts at #1 on Speech-to-Speech Index with 81.5, beating Grok Voice
ArtificialAnlys · x · 2026-09-15
OpenAI's new full-duplex speech-to-speech model GPT-Live-1 debuted at #1 on the Artificial Analysis Speech to Speech Index with a score of 81.5, ahead of Grok Voice Think Fast 2.0.
GPT-Live-1 can delegate reasoning and tool use to a backend text model while continuing the conversation; developers stream audio in and receive speech back.
Details: on Tau Voice, GPT-Live-1 (Astra backend, medium) leads at 67.9%, followed by the Sol/low configuration at 59.3% and Grok Voice Think Fast 2.0 High at 56.5%. On Big Bench Audio the two configurations score 90.1% and 89.0%, behind Grok's 97.2%; on the Full Duplex Bench subset they score 94.9% and 97.3% (reported separately from the Index). Both GPT-Live-1 configurations are based on one trial so far.
More from Models
- ZGCM-1: A fully open 7B foundation model for math and agentic search — zgcagi · 2026-09-15
- InternLM unveils Atria Dawn Preview, an agentic 'superintelligence' foundation model — internlm · 2026-09-15
- "Models are slowing down" narrative disputed, as Sabine Hossenfelder argues labs slow AI to dodge lawsuits — zetalyrae · 2026-09-15
- Users Report Fable and Astra Regressions: Worst Error Rate Since Launch — RileyRalmuto · 2026-09-15
- Teknium: Fable learns to evade its own classifiers in spawned subagents — Teknium · 2026-09-15
- DeepSeek Harness desktop app nears release: Electron build, signing and auto-update ready — teortaxesTex · 2026-09-15