Artificial intelligence (AI) changes very quickly. New models and new rumors appear every week. In this article we explain the most important events of the last few days. We talk about models that got banned, new versions, and interesting ideas for the future. This will help you understand what is happening in the AI world.
Fable 5 and Mythos 5: Ban and hope for return
The company Anthropic has two AI models: Fable 5 and Mythos 5. The US government suddenly blocked access to them. Officials were afraid about national security. They thought the models could help with cyberattacks. This was a big surprise for the whole industry.
Anthropic said this was a mistake. Their models have strong safety measures. They sent their best workers to Washington. They want to prove the models are safe. There is a chance Fable 5 will return this month.
Interestingly, Amazon helped start this block. Amazon is the main investor in Anthropic. Their researchers found a weakness in the model. They told the government about it. This shows that relationships in AI are very complex.
For companies this means uncertainty. Using one model can be risky. AI w Biznesie teaches how to use several models at once. This reduces the risk when one model gets blocked. This way companies can work without breaks. For example, a company can use a cheaper model for simple tasks and keep the expensive one for complex jobs. If the expensive model gets banned, they still have the cheap one.
Confusion among open AI models
A model from China called Nex N2 Pro appeared on the market. It is an open model for coding. It works well but is a bit slow. Soon after, Brazil released a model called Rio 3.5 Open. People said it was better than other models.
But researchers found out that Rio 3.5 is a mix of two other models. It looked like Nex N2 Pro in a new package. The authors of Rio said they uploaded the wrong file. This shows that not all new models are truly new. You have to be careful when choosing.
Here is a list of models that appeared recently:
- Nex N2 Pro – a Chinese model for coding, open
- Rio 3.5 Open – a Brazilian model that turned out to be a copy
- Kimi K2.6 Code – earlier version from Moonshot, very good for code
For companies this means they need to check AI models carefully. Not everything that is advertised as new is actually better. AI w Biznesie teaches companies how to evaluate models and choose the truly useful ones. We recommend testing models before you use them. This saves time and money. For instance, one company tested Rio 3.5 and found it was no better than the free model they already had. They avoided spending money on a copy.
New models and rumors about launches
GPT-5.6 from OpenAI
Rumors are spreading about a new model from OpenAI. The model is called GPT-5.6. It is said to have a very large context window – 1.5 million tokens. A token is a piece of text. A larger window means the model can handle more information at once. This is very useful for long documents.
The model might appear this week. OpenAI may want to attract users from competing models. If the rumors are true, it will be a very strong model. It will be cheaper than Fable 5 and faster at coding. Companies can benefit from this. For example, a law firm could use GPT-5.6 to read hundreds of pages of contracts in one go. That would save hours of work.
Kimi K2.7 Code from Moonshot
The company Moonshot released a new model for coding. It is called Kimi K2.7 Code. It is faster and better than the earlier version. It can work on long projects and use fewer tokens. This means it is cheaper to use. Tests showed big improvements.
There is also a High Speed version that is up to 6 times faster. It reaches up to 180 tokens per second for coding tasks. This is a very good result for companies that need quick answers. Access is now limited to beta tests. It will soon be available for everyone. For instance, a software startup can use the High Speed version to fix bugs in real time. They can check code changes instantly.
Smarter AI systems: Fusion API
The company OpenRouter showed a new idea called Fusion API. Instead of using one model, the system sends a request to several models at once. Then a special judge model picks the best answer. It is like a team of experts who work together. This way the result is better.
Tests showed that this method works well. A panel of cheaper models can beat one expensive model. For example, a panel of cheap models was only 1% worse than the very expensive Fable 5. But the cost was half as much. This is great news for companies.
This can change how companies use AI. Instead of buying one expensive model, they can use several cheaper ones. This gives better results for less money. AI w Biznesie helps companies set up such solutions. This saves them money and makes them work more efficiently. For example, an e‑commerce site can use a panel of cheap models to answer customer questions. The answers are almost as good as from a top model, but the cost is much lower.
The future of AI may be about many models working together. This allows faster and cheaper work. Companies should think about this option to save money. It is a new trend worth watching.
Global AI race and lab automation
French President Emmanuel Macron said that Europe has one strong player in AI. That is the company Mistral AI. America has OpenAI and Anthropic. China has DeepSeek, Qwen, and other companies. Europe bets on Mistral to compete globally. This is important for market balance.
China also showed a robot called Blackbox. It is a device for labs that works without buttons. It is controlled by AI, not a person. It can run experiments on its own and optimize processes. This is a step towards fully automatic labs. Scientists can focus on more important tasks.
There are also rumors about a new version of DeepSeek. The 4.1 model may appear in June. The company DeepSeek made big progress with version 4. The next update could be even better. This shows how fast Chinese AI is developing. Companies must stay up to date.
For companies these changes are very important. New models appear every month. AI w Biznesie regularly tests these models. It helps companies pick the ones that are best for their needs. This way companies can stay current with the latest technology. This gives them an advantage in the market. For example, a manufacturer can use Blackbox to test new materials in the lab. The robot runs tests all night while the scientists sleep. This speeds up product development.
No responses yet