Elon Musk and the New Grok 5 Model
Elon Musk surprised the world of artificial intelligence. His company xAI announced a new model called Grok 5. It has 1.5 trillion parameters.
Parameters measure how big an AI model is. More parameters mean more knowledge. The new Grok is three times bigger than the last version.
The model is already fully trained. Musk says it will reach the public in 2 to 3 weeks. This could be xAI’s biggest move in the AI race.
A company called AI w Biznesie helps businesses use similar tools. We show firms how to train AI with real work data. That makes automation more effective.
Training Data from Cursor
The most interesting part is how xAI trained Grok. The firm used a huge amount of data from a tool called Cursor. Cursor is a popular AI for writing code.
More than 67% of Fortune 500 companies use Cursor. The tool is expected to earn 6 billion dollars by 2026. Jensen Huang from Nvidia called it his favorite AI tool.
Cursor data shows how real programmers work. It records how people write, test, and fix code. It is like studying for an exam with the answer sheet.
AI w Biznesie offers training that uses similar methods. We teach companies to collect data from their own workers. Then the AI learns exactly what the firm needs.
How Grok Learns from Coders
Current AI models can write code that looks good. The real challenge is understanding complex logic. Grok learns this from Cursor data.
The data includes the questions programmers ask. It shows their fixes and debugging sessions. It also shows work with many files at once.
Someone asked Grok directly about Cursor data. The model said it contains high-quality interactions. There are commands, code context, and task logs.
Our firm AI w Biznesie offers similar solutions for businesses. We help you set up AI that learns from your team. This makes your automation tools much better.
Musk’s Plan for the AI Market
SpaceX, another Musk company, made an offer for Cursor. It wants to pay 60 billion dollars for the option to buy. If it does not buy Cursor, it will pay 10 billion dollars as a fee.
This shows how much Musk values the AI coding market. His plan has three steps. First, he secures Cursor with money. Then he feeds data to his model. Last, he launches his own agent.
The agent is called Grok Build. It started working on May 14. It is an AI tool that works in the command line. It can write code, edit files, and manage tools.
Grok Build can work with eight sub-agents at the same time. The subscription costs 300 dollars per month. For the first six months, there is a deal for 99 dollars.
AI w Biznesie can help you understand these new tools. We show you which AI agents fit your business. You save time and avoid costly mistakes.
Grok’s Position in the Market
Grok Build works with the file format used by Claude Code. This means it is compatible with a competitor’s product. xAI is building bridges to the rival ecosystem.
Even so, Grok is far behind. On the SWE-bench test, GPT‑5.5 gets 88.7% success. Claude Opus 4.6 gets 80.8%. Grok only gets 72-75%.
In companies, OpenAI has 55% of the market. Anthropic jumped from 20% to 47% in one year. Google has 39%. Grok has only 6%. Musk has a lot to catch up on.
The dates of actions are not random. SpaceX will list on NASDAQ on June 12. The offer for Cursor should be final 30 days after that. Grok V9 will come out just before the IPO.
Our team at AI w Biznesie tracks these numbers closely. We help you choose the model that gives the best results for your work.
Big AI Month in June
Musk is not the only player in June. OpenAI will release GPT‑5.6 with a huge context window. Claude Opus 4.8 appeared on Google Vertex.
Gemini 3.5 Pro is also coming out in June. The four biggest AI labs will fight. This month will be a turning point for the industry.
There is also a legal matter. xAI sent guidelines to its employees. They must limit contact with Cursor staff to a minimum.
This is a normal step during acquisition talks. Antitrust law forbids combining assets before approval. The technical partnership continues, but with distance.
AI w Biznesie can help you prepare for these changes. We offer workshops on new models and tools. Your team will stay ahead of the competition.
Revolution in Scientific Research
DeepSeek published a 46‑page scientific paper. An AI agent wrote 99% of the text. A human wrote only 1%.
The author is Deli Chen, a senior researcher at DeepSeek. He worked on models V1, V2, V3, and V4. His agent is called Delhi Auto Research Skill.
The first version of the paper was made in 76 minutes. The whole process took six days. The agent ran 108 rounds of interactions and used 648,000 tokens.
The paper has 103 verified sources. It contains seven figures and four tables. The researcher spent less than two hours of thinking on it.
Deli Chen talks about crazy inflation in scientific papers. What used to take months now takes days. The co‑authors are AI models: DeepSeek V4 Pro and GPT Image 2.
Our firm AI w Biznesie helps companies understand these changes. We teach how to use AI for research and writing. That saves your team months of work.
Five Levels of Autonomy in Research
The paper proposes a system of five autonomy levels. This is similar to the classification for self‑driving cars. Level one is auto‑complete, like GitHub Copilot.
At level two, a human defines the task and approves actions. This works like ChatGPT with tools. Level three is multi‑step operations with checkpoints.
Level four is full autonomy within a limited scope. A human gives a goal and checks the result. Level five is independent research without supervision.
Current systems work at level four. Level five is still a dream. The biggest barriers are memory and self‑evaluation without a human.
AI w Biznesie helps firms understand these levels. We advise on how to use AI safely. That way your company avoids costly errors.
Chinese Model Surprises the World
Alibaba showed the Qwen 3.7 Max model. It took fourth place worldwide in coding. It beat GPT‑5.5 and Gemini 3.5 Flash.
This is the first Chinese model in the top coding rankings. Only Claude Opus 4.7 and 4.6 are better. Alibaba is the only Chinese maker in the top five.
Developers tested Qwen in practice. One asked it to write an AI for the game Tetris. Qwen not only won, but cost only 1.32 dollars.
Another developer built a 3D model of the universe with Qwen. The results were very good. The model also created racing games with engine sounds.
Why Qwen Is So Good
Qwen 3.7 Max was designed as a model for agents. It can work on its own for 35 hours. During that time, it makes more than 1,000 tool calls.
After 30 hours of work, the model was still focused. It did not lose the goal or forget earlier decisions. It did not fall into error loops.
Most models break down on such long tasks. Qwen keeps its consistency for many hours. This is a strong signal for the competition.
The training method may explain the success. The same code was tested in different environments. The model learns general patterns, not shortcuts for one system.
The AI coding market is changing very fast. Grok gains data from programmers. Qwen shows the power of Chinese models. DeepSeek proves AI can write scientific papers.
Companies must follow these changes. AI w Biznesie offers training on new technologies. We help you pick the best models for your business.
June 2026 will be a breakthrough. The four biggest AI labs will release new models. The race for the best coding AI is speeding up.
For businesses, this is great news. More competition means better and cheaper tools. Automating business processes will become easier.
If you want to learn more, contact us. We will help you choose the right AI solutions for your company. Subscribe to our channel for more updates.
No responses yet