SpaceXAI launched Grok Voice Think Fast 2.0, an AI model that succeeds the Grok Voice Think Fast 1.0 launched earlier this year. The company said in a release that this is their most intelligent voice model yet, building on its predecessor with meaningful gains in speech reasoning, conversational ability, and tool use reliability.
SpaceXAI highlights the model’s transcription capabilities, reasoning efficiency, and conversation capabilities. Grok Voice Think Fast 2.0 outperforms even dedicated, state of the art transcription models when it comes to accuracy.
READ: Elon Musk pledges to make X open source after Grok backlash (July 15, 2026)
Grok Voice Think Fast 2.0 outperforms even dedicated, state of the art transcription models when it comes to accuracy, the company said. It also mentioned that Grok Voice Think Fast models have a unique characteristic: they reason through queries while speaking. Reasoning in parallel with speech makes the model substantially smarter than other speech-to-speech models with no impact on latency. The tech giant also mentioned the model has been trained to be a better conversationalist.
SpaceXAI said that during the A/B testing of the Grok Voice Think Fast 2.0 model on Starlink, the tech giant witnessed “a significant increase” in sales conversion rate and support containment rate. The company also compared Grok Voice Think Fast 2.0’s benchmark results against Grok Voice Think Fast 1.0, OpenAI’s GPT-Realtime-2.1, and Google’s Gemini 3.1 Flash models, and claimed that the Grok Voice Think Fast 2.0 model outperformed the three models on most tests.
SpaceXAI said Grok Voice Think Fast 2.0 is priced at $0.08/min of audio.
READ: Elon Musk threatens lawsuit over Ro Khanna’s DOGE criticism (June 22, 2026)
This comes shortly after the company’s CEO Elon Musk announced a tentative launch timeline for the next two versions of xAI’s Grok artificial intelligence model, saying Grok 4.6 is expected around Aug. 7, and a significantly larger Grok 4.7 model will follow a few weeks later. Musk shared the timeline in a post on X, saying Grok 4.6 will be a 1.5 trillion-parameter model with improvements to supervised fine-tuning and reinforcement learning.
Musk’s post indicates that Grok 4.6 will focus on improvements to the model’s post-training processes. SFT, or supervised fine-tuning, involves training a model using curated examples, while reinforcement learning is used to optimize model behavior based on feedback or defined objectives. The larger Grok 4.7 model, meanwhile, is expected to increase the model’s parameter count to 2.1 trillion, compared with 1.5 trillion parameters for Grok 4.6. Musk said the larger model would be slower to serve, but would offer improved token efficiency.