The cost-effective AI model 'Grok 4.3' has been released, along with 'Custom Voices,' a speech synthesis function that can clone human voices in under two minutes.

xAI, the AI company led by Elon Musk, has released the API for its AI model ' Grok 4.3 .' Grok 4.3 is being promoted as a cost-effective AI model. They have also released ' Custom Voices, ' a speech synthesis function with high voice fidelity.

Grok 4.3 | xAI Docs

Grok 4.3 is now live on the xAI API. It's our fastest, most intelligent model to date.

It tops the @ArtificialAnlys leaderboards in agentic tool calling and instruction following, and ranks #1 in @ValsAI enterprise domains like case law and corporate finance.

Grok 4.3 supports… pic.twitter.com/83NiWoFDY2

— xAI (@xai) May 5, 2026

Grok 4.3 is an AI model that supports input of 1 million tokens. In intelligence performance tests conducted by the third-party organization Artificial Analysis, it outperformed Meta's Spart. However, it falls short of cutting-edge models such as GPT-5.5, and is inferior to Chinese-made open models such as Kimi K2.5 and MiMo-V2.5-Pro.

The graph below shows processing time per token on the horizontal axis and intelligence performance on the vertical axis. The purple dots enclosed in the red box represent Grok 4.3, which is capable of faster processing compared to Gemini 3.1 Pro and others.

The graph below shows the cost-effectiveness of various AI models. The horizontal axis represents the price per million tokens, and the vertical axis represents intelligence performance. It can be seen that Grok 4.3 is an inexpensive yet high-performance AI model.

The API usage fee for Grok 4.3 is $1.25 (approximately 195 yen) per 1 million input tokens and $2.50 (approximately 391 yen) per 1 million output tokens. Eric Jiang , a researcher in xAI, highlights Grok 4.3's high tool calling performance and cost-effectiveness.

When training Grok 4.3, we spoke directly with devs and businesses to understand what they actually needed: a model that's fast, affordable, and great at tool calling. The result is a daily driver that doesn't just look good on random benchmarks, but is actually useful in the… pic.twitter.com/z02HFXCZn5

— Eric Jiang (@veggie_eric) May 1, 2026

Furthermore, Aditya Gupta, also a researcher in xAI, has revealed that a larger model than Grok 4.3 is currently being trained.

Great work by the team.

4.3 pushes the pareto (+4pts on AA and +13 for the same size as 4.20 ~0.5T) w/ knowledge work and coding improvements.

ps: larger models being assembled in model factory. https://t.co/fHKQgQW9JY

— Aditya Gupta (@adityagupta) May 1, 2026

In April 2026, Musk revealed that he was training seven AI models: 'Imagine version 2,' 'two AI models with 1 trillion parameters,' 'two AI models with 1.5 trillion parameters,' 'an AI model with 6 trillion parameters,' and 'an AI model with 10 trillion parameters,' and that it would take two months to complete the training.

SpaceXAI Colossus 2 now has 7 models in training:

- Imagine V2
- 2 variants of 1T
- 2 variants of 1.5T
- 6T
- 10T

Some catch up to do.

— Elon Musk (@elonmusk) April 8, 2026

Furthermore, xAI has added a feature called 'Custom Voices' to its Grok Voice API, a speech synthesis API.

Custom Voices and Voice Library | xAI
https://x.ai/news/grok-custom-voices

Custom Voices is a feature that can recreate a voice from a short audio recording, and it's advertised as being able to 'clone your voice in under two minutes.' The following post includes a video recording of a real human voice and a cloned voice, demonstrating the high quality of the cloned voice.

Two voices. One human. One AI. Can you guess the AI clone? 👇

Voice cloning, rich with natural emotion, is now live on the Grok Voice API. https://t.co/EjxjXstoiL pic.twitter.com/OkGeua3H1g

— xAI (@xai) May 4, 2026