SpaceXSI Grok Voice 3.0 Ranks First on Artificial Analysis Speech-to-Speech Index

SpaceXSI released Grok Voice 3.0 today, a new iteration of the speech-to-speech model powering its Voice Agent API. According to SiliconANGLE, the independent testing firm Artificial Analysis ranks this version first among comparable models, marking a significant shift in the competitive landscape for voice AI.

The Artificial Analysis Speech-to-Speech Index rated Grok Voice 3.0 at 86.9, based on reasoning tests, task success rates, and listener votes in blind head-to-head calls. This score places it ahead of OpenAI Group PBC’s GPT-Live-1, which scored 83.2, and Google LLC’s Gemini 3.8 Live, which landed at 82.6. As Threat Times reported on October 15, 2026, the previous version also topped this index, maintaining consistent performance leadership.

Performance and Cost Metrics

Beyond the primary index, the new Grok model demonstrated strong capabilities in handling natural conversation dynamics. The Full Duplex Bench, which evaluates how models manage pauses and interruptions, awarded the new Grok model a 97% score. In contrast, OpenAI’s model maintained a narrow lead on the Tau-Voice test, which simulates support calls with heavy accents and background noise, scoring 74.5% against Grok’s 71.6%. However, SpaceXSI stated that its model delivered first audio approximately 40% faster on that specific test.

Cost efficiency also favors the new release. A task in the Artificial Analysis Big Bench Audio run cost about 2.5 cents to complete on Grok Voice 3.0. This is lower than the 3 cents charged for GPT-Live-1 and significantly less than the 5.5 cents for OpenAI’s GPT-Realtime-2.1. These figures suggest a competitive advantage for businesses prioritizing both performance and operational expenditure in voice agent deployments.

Enterprise and Consumer Adoption

SpaceXSI is actively pitching the model to businesses managing customer support and sales calls. Outside of Artificial Analysis’ testing, Grok Voice 3.0 led for accuracy on ServiceNow Inc.’s open-source EVA-Bench. It scored 74% across more than 200 enterprise scenarios, a metric that combines accuracy with customer experience. No other model achieved a higher combined score, according to the source material.

In production environments, Tesla Inc. vehicles currently process millions of Grok Voice conversations daily. Drivers use the system for routine tasks such as checking calendars and answering emails. More complex requests, like booking dinner, are handled by Grok Bot, which utilizes the new voice model. At Starlink, the SpaceX satellite internet business, the model answers roughly 20,000 customer calls daily in several languages. SpaceXSI reported that about 70% of these calls are resolved without human intervention, and one in four sales calls results in a purchase before the caller disconnects.

What should you do?

Security and technology professionals should monitor vendor announcements regarding API integrations for voice AI models. As companies adopt these systems for customer support, evaluating the security posture of the underlying voice data pipelines becomes critical. Organizations should review how voice data is stored, processed, and shared by third-party providers like SpaceXSI.

What to do and how to stay safe: SpaceXSI

• Audit current voice agent integrations to ensure they comply with internal data privacy and security policies.
• Monitor vendor documentation for changes in data handling practices as new model versions are released.
• Evaluate the accuracy and bias metrics of voice models in your specific operational context before full deployment.
• Implement logging and monitoring for voice interactions to detect anomalies or unauthorized access attempts.

Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

How does Grok Voice 3.0 compare to OpenAI and Google models?

Artificial Analysis ranks Grok Voice 3.0 first with an 86.9 score, ahead of OpenAI’s GPT-Live-1 at 83.2 and Google’s Gemini 3.8 Live at 82.6.

What is the cost of using Grok Voice 3.0?

A task in the Artificial Analysis Big Bench Audio run costs about 2.5 cents on Grok Voice 3.0, compared to 3 cents for GPT-Live-1 and 5.5 cents for GPT-Realtime-2.1.

Is Grok Voice 3.0 currently in use?

Yes, Tesla vehicles use it for millions of conversations daily, and Starlink uses it for approximately 20,000 customer calls a day.