Editorial illustration for SpaceXAI Grok Voice 3.0 Ranks #1 on Artificial Analysis Speech-to-Speech Index
Grok Voice 3.0 Tops Speech-to-Speech Benchmark
SpaceXAI pushed Grok Voice 3.0 live this week, and Artificial Analysis already has it sitting at the top of its Speech-to-Speech Index with a score of 86.9, the highest mark any model has posted on that leaderboard. The model is exposed as grok-voice-3.0 on the /v1/realtime WebSocket endpoint, and the grok-voice-latest alias now points straight to it, so agents built on the previous version switched over without anyone touching a line of code. Pricing stays at $0.08 per minute, the same as Grok Voice 2.0, and the API events haven't changed either.
SpaceXAI hasn't released parameter counts or architecture specifics, and there's no open-weight or quantized version floating around, this is strictly a hosted API play. Reasoning runs at "high" effort by default, with an option to drop it to "none" when speed matters more than depth. Barge-in is handled through server-side voice activity detection, and tool use, collections search, web search, X search, remote MCP servers, custom functions, all run inside the session itself.
The Index score comes from four equally weighted components:
SpaceXAI just released Grok Voice 3.0. Grok Voice 3.0 is a SpaceXAI speech-to-speech model that listens, reasons, calls tools and speaks. It now ranks first on the Artificial Analysis Speech-to-Speech Index with a score of 86.9.
Why this matters
The 86.9 score matters less than the price tag staying flat at $0.08 per minute. xAI could have charged a premium for the top spot on Artificial Analysis's index, beating GPT-Live-1's 83.2, but chose not to, which tells us this is a retention play, not a margin play. For developers already building on Grok Voice 2.0, the unchanged endpoint and events mean a drop-in upgrade with no migration cost, which is the real story here, not the leaderboard.
We'd flag the obvious limitation before anyone gets too excited: this is hosted-only, no open weights, no local or quantized builds. If your product depends on on-device inference or you need to self-host for latency or compliance reasons, Grok Voice 3.0 doesn't solve that problem no matter how it scores on Big Bench Audio or τ-Voice agentic tasks. Founders evaluating voice agents should treat the Index score as one input, not a verdict, and test task success rate against their own use case. Worth watching: whether Google or OpenAI respond with pricing moves rather than just benchmark pushes.
Related Reading
Research & Benchmarks Microsoft AI Chief: "If It Isn't Safe We Shouldn't Build It
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Related Topic Anthropic sues Trump as xAI restarts AI coder push, Iran releases AI fakes
Open Source
Related Topic Musk merges xAI with SpaceX, forming USD 1.25 trillion private giant
Business & Startups
From Archives 17 Organizations Use AI to Accelerate Scientific Breakthroughs
From Archives Google: self-improving AI far off as fully delegated code changes rose to 11.9%
From Archives SpaceX Bundles X Premium, Grok AI, and Cursor for Developers in One Subscription
Business & Startups
From Archives Shivon Zilis Comments on Split From Elon Musk
Business & Startups
Common Questions Answered
What score did Grok Voice 3.0 achieve on the Artificial Analysis Speech-to-Speech Index?
Grok Voice 3.0 achieved a score of 86.9 on the Artificial Analysis Speech-to-Speech Index, making it the highest-ranked model on that leaderboard. This score surpasses the previous top performer, GPT-Live-1, which scored 83.2.
How does Grok Voice 3.0 handle the migration from the previous version?
Grok Voice 3.0 uses the same /v1/realtime WebSocket endpoint as its predecessor, with the grok-voice-latest alias now pointing to the new version. This means developers can upgrade without modifying any code, as it functions as a drop-in replacement with no migration costs.
What pricing model did SpaceXAI maintain for Grok Voice 3.0?
SpaceXAI kept the pricing for Grok Voice 3.0 at $0.08 per minute, the same rate as the previous version. Despite achieving the top ranking on the Artificial Analysis index, the company chose not to implement a premium pricing tier, indicating a focus on customer retention rather than margin expansion.
What capabilities does Grok Voice 3.0 possess as a speech-to-speech model?
Grok Voice 3.0 is a speech-to-speech model that can listen, reason, call tools, and speak in response. These capabilities enable it to function as a comprehensive voice agent for developers building conversational applications.
Further Reading
- Grok Voice 3.0 from SpaceXSI outscores OpenAI and Google on speech-to-speech index - SiliconANGLE
- Elon Musk takes aim at the customer service industry - Yahoo Finance
- Announcing the Artificial Analysis Speech to Speech Index - Artificial Analysis
- Speech Reasoning Benchmarking Methodology - Artificial Analysis
- Grok Voice 3.0 Launches With Improved Call Handling and Sub-Second Latency - iClarified
