The large model arms race has entered the second half, the era of applications is here

In the field of AI, two leading model companies, Anthropic and Google, released their latest models yesterday. Anthropic's Claude Mythos and Fable5 are more suitable for running extremely complex long tasks, while also introducing the concept of safety guardrails. This means that when you use the latest model, Mythos, to access areas the model considers dangerous, its capabilities will be automatically downgraded to prevent misuse by malicious actors.

On Google's side, they also released the new generation of Google Translate, Gemini 3.5 live translate. The biggest feature of this model is its real-time mutual translation capability across multiple languages (over 70 languages). This essentially means that basic simultaneous interpretation work won't be as necessary in the future.

More importantly, this large model can be integrated with relevant apps. For example, in the case study introduced by Google, the Southeast Asian ride-hailing app Grab has already embedded this latest model. For instance, when we Chinese use Grab to book a ride abroad, it's very likely that an Indian driver will accept the order. At this point, the model comes into play. You can communicate location details with the Indian driver over the phone in Chinese. You speak your language, they speak theirs, but everything is translated into each other's native language in real-time, enabling barrier-free, almost delay-free, smooth communication.

Therefore, from the current development of AI, while iterative upgrades of the models themselves are certainly important, there is an increasing focus on the seamless integration and empowerment of the models with real-world applications.

(Not for investment purposes)

$Alphabet(GOOGL.US) $Direxion Daily GOOGL Bull 2X Shares(GGLL.US)  $ICBCCS CSI STAR&CHINEXT Artificial Intelligence ETF(588430.SH)