Deepseek can be installed on a PC or a Mac locally so that all the queries that are made remain in the machine. It is another of the experiences that has left the world speechless by not demanding so many resources and bringing the capacity of GPT-4o in a ‘distilled’ model that offers great flexibility in its use. You can also mobile as Samsung Galaxy S24 ultra himself.
As is the case when Deepseek is installed on a PC or Mac, according to the major language model that is used, its effectiveness will be greater or less, not only in the answers that you can give to the consultations, but in the speed of developing them .
There is another important point in Depseek models that can be loaded on a high -end mobile, and it is that Optimized versions of Deepseek R1 They have the objective of being installed in machines with less resources.
And the better the PC or mobile, it will offer a better experience, so you have to try to choose the model that will be installed well, since if it is not done properly, the speed of your responses can be drastically reduced as well as your precision.
The tests done by the Spanish – the free android almost pass more through an experiment, since the model that has been able to offer more precise answers has need more time to process them (As with Depseek-R1-Distill-Qwen-14B-IQ2_M) and being Deepseek R1, it will show in text all the logic of each of its ‘deep thoughts’ to give the final answer.
Asking for the torrijas recipe to Depseek in a Samsung Galaxy S24 Ultra
The free android
It really is one Sample of the efficiency of this major language model that competes against GPT-4o and that can be installed on an Android mobile. Something practically unthinkable days ago and that is already available in our hands, although in a limited way.
The Deepseek code repository of its different models is available from Hugging Facea platform that is becoming one of the nodes to share tools, models and data sets all related to generative artificial intelligence.
Necessary requirements
The compilations we will use have been published just days ago As has happened with Deepseek R1, which is also available from this repository, although it needs a huge number of resources that can offer a PC, even if it is the best gaming.
THE DEPEEEK-R1-DISTILL-QWEN-14B-IQ2_M He needed 724 ms per Token with 1.38 tokens per second to give all the details of how to make a recipe for toast and with certain incorrect words. It is slow, but it is also a matter of trying other submodels to accelerate the answer and verify that their answers are more or less precise.
Samsung’s mobile is one of the best current mobiles for its specifications and certain requirements are recommended to execute some of Depseek R1 models that can be downloaded from Hugging Face:
- Chip: Snapdragon 8 Elite, Snapdragon 8 Gen 4 / Gen 3 and some Snapdragon 7+ Gen 3 as well as those of Mediatek as Dimensity 8400 or similar.
- RAM: At least 8 GB for the models 8b / 14b of Deepseek.
- Storage: At least 12 GB free.
How to install Depseek on a mobile
To install Depseek on the mobile The installation of Pocketpal AI, an app is needed Open Source which facilitates all the steps to download a major language model and then load it to operate with the chatbot.
It is downloaded From the Google Play Store In Android and is also available for iOS on the iPhone From the app store. Once the app is open, you just have to click on the “Go To Models” button, and then touch on “+” and choose “Add from Hugging Face”.
Deepseek R1 submodels
The free android
Here the complete repository of this platform will appearso the search engine should be used to type “Depseek”. Among the entire results list is recommended to use a larger model such as 8B, since 1.5b or 7b are faster, but more imprecise in their answers.
Each model then has its submodels which vary in capacity and file size to download. The greater, the greater the resources, so according to the mobile, we will have to prove one or the other, we will use Deepseek-R1-Distill-qwen-7b-Q3_K_L with 4.09 GB and 7.62b of parameters.
The model to be installed and the adjustment to be made
The free android
We download the submodel and at the end we go to the “Models” section From the main menu. It is pressed on “settings” in the downloaded model and on the next screen you have to configure “N Predict” with 4096 tokens to generate text properly.
He returns to the previous screen and is played in “Load” To start the selected model and so you can move on to a conversation window in which you can consult, see the answers and even start another chat with AI.
Deepseek will return the answers with “deep thought” by marking it as “Thinking …”. When you finish it will give the answer to continue with the conversation. We remember again that once the session is finished, it is recommended to download with “Unload” on the model from the previous screen.
