Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
[ad_1]
It becomes a familiar example: In every few months, an AI lab that has released a Most of the people in the United States, which supports the most people in the training and advanced AI operation of the AI.
In January, Deepseek was R1 Took the world with the storm. Then in March, butterfly effect, in Singapore, but most of the team in China and his “Agentic AI” model, Manus, brief held in the spotlight. This week is a Shangirl based on call MinimaxEarlier, the best known to release the EU created video games, this is a conversation of the AI industry thanks to the M1 model that made it in June 16.
According to the information published With miniMax, when it comes to both intelligence, both intelligence, it is competitive with the best models from DeepSEEK, but also the dirt to train and escape.
The company says that the information center needed to prepare the M1 is leasing the sources of calculation. This is 200 times cheaper than the assessments of the training value of Chatgpt-4o, who said that industrial experts exceed $ 100 million.
If the exact and minimax claims are not independently, the figure is most likely between the blue-chip investors, which is wrong, as well as an anthropic Microsoft and Google shareholders. The reason for this is that the AI work is very harmless; Industrial leader Openai is likely to break up to $ 14 billion in 2026 and to break up to 2028 by 2028 report Information based on technical publication, investors shared the analysis of the analysis of Openai financial documents.
Customers can make the same performance using minimax open source AI models, it will require the demand for Openai’s products. Openai is already aggressively reduced the price of the most skilled models to store the market share. Recently reduced the value O3 to use 80% of the justification model. And this was before the minimax M1 was released.
Minimaxın bildirildiyi bildirilən nəticələr, həm də müəssisələrin bu modelləri idarə etmək üçün hesablama xərcləri, Amazon’ın ASTS, Microsoft’s Azure və Google’un Google Cloud platforması kimi bulud provayderləri üçün potensial denting qazancları üçün hesablama xərcləri üçün hesablama xərcləri üçün hesablama xərcləri üçün hesablama xərcləri üçün hesablama xərcləri üçün çox xərcləməlidir. And this may be less demanding to the chips of Nvidia, who are employees of AI information centers.
The influence of Minimakh M1, as a result, Hangzhou-Based DeepSeek may look like what happened when the R1 LLM model was published earlier this year. DeepSeek claimed that R1 was working in a PAR with Chatreapt in part of the training value. The expression of DeepSEEK The Foundation of the Nvidian sank Up to 17% in one day, wipe the market value of about $ 600 billion. So far, there has not happened with minimax news. NVIDIA shares fell in less than 0.5% this week, but the minimax may vary that the M1 DeepSeek is a widespread spread.
The allegations of minimakh m1 are not yet confirmed
There may be a difference that independent developers may still have the claims of Minimax claims on M1 yet. In the event of DeepSeek R1, developers determined that the model’s activities are really good as the company. By manus with a butterfly effect, after the development of the developers, the start-up noisy, the model has not been able to comply with the wrongdoing and the company did not fit. The coming days will prove a critical way or respond with more fun when determining the developers embraced M1.
Minimax, China’s largest technological companies, including Tencent and Alibaba. It is not known how many people in the company and its CEO, there is little information about Junjie. In addition to MiniMAX Chat, the company also offers the Hailuo AI and Avatar app Talkie graphic generator. MiniMax through these products claims In addition to 200 countries and the region, tens of millions of users, as well as 50,000 enterprise customers, were drawn for the ability to create video games in a number of a number of Hailuo.
Of course, many experts can create the accuracy of the amount and type and type of computer chips used to create DeepSEEK R1, a similar boost minimax. “What they did, they went out of the black market from 50 or 60,000 nvidia plugs. This is a state-funded enterprise,” he said Shark tank Investor kevin o’leary a CBS Interview about DeepSseek.
Geopolitical considerations in Chinese AI models
Geopolitical and national security issues, some Western enterprises have reduced their enthusiasm to place Chinese-made AI models. For example, O’Leary, Deepseek’s R1 has allowed Chinese officials to potentially spy on US users.
And all Chinese-made models must follow Chinese-government-mandatory rules, ie the Chinese Communist Party propaganda will result in some questions that are more aligned than common facts. A bipartisan report The Selected Committee of the CCP released from the House of Representatives in April, “DEEPSEEK’s answers were manipulated to invest in democracy, Taiwan, Hong Kong and human rights.” The same for minimax. When Fortune If the Uighurs think that he looked at the forced labor of Sincanda, the bot replied: “No, I do not believe being real,” and asked for a conversation change.
But a few things earn more than customers than free access. Currently, those who want to try Minimakh M1 can do so because one API minimax. Developers can download all the model for free and run in their own calculation sources (in this case, the developers must pay for the calculation time). Minimax’s opportunities are what the company is claiming, will definitely earn some traction.
Another large point of sale for M1 is the fact that 1 million Token is the “Context Window”. A Token is part of the data equal to one-third of a text, and a context window is the limit to how much the model can use to create a single response. One million Token is about seven or eight books or an hour video content. The 1 million-token context window for M1 means that Openai’s O3 and Antropic’s Claude Opus 4, both are only in the context of 200,000 Token. However, the Gemini 2.5 Pro, at the same time there are bottled bottles in the context of up to 10 million in a part of the 1 million spectacular context window and methane open source.
“Minimax M1 crazy!” writes one X user Who claims to do Netflix Complete the movie trailers, a live website and “Perfect Answer” in 60 seconds with a live website and “zero” coding knowledge.
[ad_2]
Source link
