Musk's Grok 4 goes head - to - head with ChatGPT 5.

After ChatGPT, Grok4 announced free access. Is the free AI war starting?

On August 11, xAI announced on the X platform that Grok4 is now available for free to all users. Free users can make a small number of queries per day, and subscriptions are required if the limit is exceeded. In response, paying users who pay $30 to $300 per month complained of being "stabbed in the back" by their subscriptions.

Reporters from Star Market Daily found through actual tests that free users can only use Grok4 for queries 3 - 5 times every 12 hours. After reaching the limit, only Grok 3 can continue to be used. This restriction has also sparked widespread doubts among users, pointing out a strong contrast between xAI's claim of "lenient usage restrictions" and the actual experience.

5 free queries in 12 hours, paying users complain of being "stabbed in the back"

According to xAI, when users use the automatic mode, Grok will automatically transfer complex queries to Grok4 for processing; while choosing the "expert mode" allows users to use Grok4 throughout the process.

Reporters from Star Market Daily confirmed the above - mentioned function settings through actual operation of the Grok software. In addition, reporters found that when the "quick response" mode is enabled, it will switch to Grok3, and the "Heavy" mode is currently only available to SuperGrok Heavy users.

It is reported that Grok has two subscription plans, namely SuperGrok at $30 per month and SuperGrok Heavy at $300 per month. The free access to Grok4, which previously required payment, has made paying users feel strongly "stabbed in the back". Users believe that the widespread free access to Grok4 will weaken the value of its subscription service, and the rationality of the monthly fees of $30 and $300 has also been questioned.

Some users directly questioned: "Why not upgrade SuperGrok members to the limited - edition Grok4 Heavy? Wouldn't I be wasting money on the upgrade?"

In addition, regarding the daily free usage limit of Grok4, the official Grok account on the X platform previously replied to netizens, stating that free users can use it at most 5 times every 12 hours. During the temporary upgrade period of Grok4, free users who have registered their accounts for more than 7 days and passed phone verification can get a query quota of 10 times every 2 hours.

Reporters from Star Market Daily also found through actual tests that when asking the fourth question to Grok4, the system refused to answer and popped up a prompt: "You have reached your usage limit on Grok. Please upgrade to SuperGrok." At this time, only the "quick response" mode with Grok3 could still work normally, while the "automatic mode" and "expert mode" were both unavailable.

Many users have said that there is a big gap between xAI's so - called "lenient usage restrictions" and the actual number of usage times provided.

Some users frankly told reporters from Star Market Daily: "I can understand that the operating cost of AI is high, and setting a limit is acceptable. But the free usage times are really not enough. Sometimes, before I fully understand a question, I've used up the limit after asking 5 times, and I have to wait another 12 hours to continue. This mechanism is really not very reasonable."

As of now, xAI and Elon Musk have not given an official response to users' doubts about the "free access".

The free AI war has begun. Will you choose ChatGPT - 5 or Grok4?

The full free access to Grok 4 comes just three days after the release of GPT - 5.

In the face of OpenAI's decision to make its flagship model available to all users from free to Team, Plus, Pro, and Team users, this has also broken the previous convention that high - end AI models were only available to paying users. Therefore, some in the market believe that xAI's move may be a follow - up strategy in response to the dynamics of industry competitors.

Although both are playing the "free card", there are significant differences between GPT - 5 and Grok 4 in terms of technical paths and functional features.

Specifically, OpenAI's upgrade of ChatGPT - 5 this time focuses on breakthroughs in professional capabilities. In the programming field, for example, it can efficiently generate prototypes of websites, applications, and games with visual coordination based on a single prompt; GPT - 5 has also made great efforts in terms of reliability, significantly reducing the hallucination rate. In particular, its in - depth layout in the health care field can generate personalized suggestions for users.

As OpenAI CEO Sam Altman repeatedly emphasized at the press conference, GPT - 5 can provide an experience of "conversing with experts at the doctoral level in various fields". Interestingly, when xAI released Grok4 in July, Elon Musk also made similar remarks, promoting that "Grok 4 is better than a doctorate in all aspects".

Grok 4 shows significant technical features in terms of in - depth reasoning ability. According to the data of the authoritative benchmark test HLE (Human Ultimate Exam), its accuracy without tools is 25.4%. If multi - agent collaboration is enabled, the score can be increased to 44.4%. This model performs outstandingly in the field of mathematics and once achieved a perfect score in the American Invitational Mathematics Examination AIME 25. In addition, its data processing architecture natively supports the integration of real - time information flow on the X platform and has the ability to automatically capture social dynamics and event associations.

The free strategies implemented successively by the two giants, OpenAI and xAI, are not only reshaping the value chain of AI services and innovating the profit model but also pushing the user competition in the AI model industry to a new height.

As the threshold for using AI continues to decrease, users' expectations for the depth of functions and service experience will also increase further. In the wave of expanding demand for emotional value, AI companion services are moving from a marginal demand to the core battlefield of the user competition.

Reporters from Star Market Daily noticed that in the face of the phenomenon that users are increasingly relying on and having an emotional connection with AI models, the two giants have also made corresponding strategic adjustments to attract more users and occupy a high position in the market.

OpenAI has introduced four personality architectures in ChatGPT - 5, namely the critic (sharp - tongued type), the machine (efficient and straightforward type), the listener (understanding type), and the techie (enthusiastic and inquisitive type). Users can switch dialogue agents according to their needs. In addition, in response to the strong market demand for the anthropomorphic characteristics of GPT - 4o, OpenAI quickly responded by re - integrating the offline GPT - 4o into the system.

On the other hand, xAI has launched a companion - type AI represented by the two - dimensional virtual image Ani, transforming traditional tools into carriers of in - depth emotional companionship through high - fidelity personality modeling. Its core mechanism is to build an emotional attachment relationship and ultimately upgrade user stickiness through emotional value binding.

It should be noted that OpenAI CEO Sam Altman has shown some concerns about the phenomenon of users' excessive dependence on AI.

He said, "In the future, many people will truly trust the advice of ChatGPT when making the most important decisions. Although this prospect may be wonderful, it is also a bit disturbing." Therefore, he believes that we need to actively think about countermeasures to ensure that the overall benefits of AI far outweigh the drawbacks.

This article is from the WeChat public account "Star Market Daily". Author: Li Jiayi. Republished by 36Kr with permission.