Your voice model probably can't answer a phone, and the docs won't tell you

Description and use

<p>Every comparison of realtime voice models measures the same things. Latency, benchmark scores, price per minute. All useful. None of them answer the question that actually blocks a phone deployment, which is whether the thing can pick up a call at all.</p> <p>We support GPT-Live-1, Gemini Live and Grok Voice, so we went looking. The answer turned out to be less about capability and more about what each vendor is willing to put in writing.</p> <h2> Sockets are not phone lines </h2> <p>Here is the split as the vendors document it today.</p> <p>OpenAI publishes a SIP connector for GPT-Live-1.

Problem

<p>Every comparison of realtime voice models measures the same things. Latency, benchmark scores, price per minute. All useful. None of them answer the question that actually blocks a phone deployment, which is whether the thing can pick up a call at all.</p> <p>We support GPT-Live-1, Gemini Live and Grok Voice, so we went looking. The answer turned out to be less about capability and more about what each vendor is willing to put in writing.</p> <h2> Sockets are not phone lines </h2> <p>Here is the split as the vendors document it today.</p> <p>OpenAI publishes a SIP connector for GPT-Live-1.

Outcome

<p>Every comparison of realtime voice models measures the same things. Latency, benchmark scores, price per minute. All useful. None of them answer the question that actually blocks a phone deployment, which is whether the thing can pick up a call at all.</p> <p>We support GPT-Live-1, Gemini Live and Grok Voice, so we went looking. The answer turned out to be less about capability and more about what each vendor is willing to put in writing.</p> <h2> Sockets are not phone lines </h2> <p>Here is the split as the vendors document it today.</p> <p>OpenAI publishes a SIP connector for GPT-Live-1.

Evidence

Your voice model probably can't answer a phone, and the docs won't tell you

Source