Sep 22, 2025 12:10:00

Introducing 'Grok 4 Fast,' a high-performance AI that can be used for free on X, with search performance higher than ChatGPT and Gemini and available to free X users

xAI, an AI company led by Elon Musk, announced the low-cost, high-performance AI model Grok 4 Fast on September 19, 2025. Grok 4 Fast is being promoted as a low-cost model with the same performance as Grok 4, and is rated as the best current model in terms of search performance, beating ChatGPT and Gemini. The beta version of Grok 4 Fast has also been released via X (formerly Twitter), making it easy for free users to use.

Grok 4 Fast | xAI

https://x.ai/news/grok-4-fast

Grok 4 Fast's context window is 2 million tokens, making it capable of handling large amounts of text data. It also employs an integrated architecture that combines 'reasoning,' which requires long periods of thought, with 'non-reasoning,' which is specialized for quick responses, in a single model, and automatically selects which inference mode to use based on user input.

The graph below shows the cost (horizontal axis) and benchmark score (vertical axis) of well-known AI models. Grok 4 Fast achieves comparable performance with Grok 4, but with a score 46 times lower.

Below is a table summarizing the benchmark scores for 'Grok 4 Fast,' 'Grok 4,' 'Grok 3 Mini,' 'GPT-5 (High),' and 'GPT-5 Mini (High).' Grok 4 Fast performed better than Grok 4 in some tests and recorded scores equivalent to GPT-5 (High) in several tests.

In tests focused on the ability to search and organize information on the web, the Grok 4 Fast outperformed the Grok 4 in every test.

In

the Search Arena , where humans anonymously evaluate the search performance of AI models, Grok 4 Fast achieved the highest score of all AI models, beating OpenAI's o3 search and Google's gemini-2.5-pro-grounding. The codename used during testing was 'menlo.'

Grok 4 Fast is available to free X users via grok.com and the Grok Chat feature in X. I gave it a try.

First, open the Grok chat screen on X and select 'Grok 4 Fast' in the model selection field at the top of the screen.

All you have to do is type your question into the chat box and send it. When I asked how to solve the problem of strange colors when recording HDR sources with OBS Studio, they gathered information from various websites and came up with a solution.

Grok 4 Fast suggested the following solution: 'Change the color format and color space settings in OBS Studio.'

When I asked, 'I have an RTX 5070ti graphics card. Which is better, AV1 or HEVC?', it recommended AV1 but also explained the situations in which HEVC would be better. Even when I asked a rough question in natural language, it was able to accurately guess what I was asking.

When asked, 'What bitrate should I set it to?', it doesn't simply give a recommended bitrate, but instead provides useful information such as, 'For recording purposes, fixed bitrates should be avoided in the first place.'

It also provided links to sources of information on the web.

Grok 4 Fast is available for unlimited use by free users on grok.com. It is also available via API , with the following pricing:

	Less than 128,000 tokens	Over 128,000 tokens
Input Token	$0.2 (approximately 30 yen) per 1 million tokens	$0.4 (approximately 59 yen) per 1 million tokens
Output Token	$0.50 (approximately 74 yen) per 1 million tokens	$1 (approximately 148 yen) per 1 million tokens
Cached Input Tokens	$0.05 (approximately 7.4 yen) per 1 million tokens