We introduced our first-generation reasoning model DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero is a model trained through large-scale Reinforcement Learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrating exceptional performance in reasoning.
With the emergence of RL, DeepSeek-R1-Zero naturally exhibited many powerful and interesting reasoning behaviors. However, DeepSeek-R1-Zero faced challenges such as endless repetition, poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduced DeepSeek-R1, which integrates cold start data before RL. DeepSeek-R1’s performance in mathematics, coding, and reasoning tasks is comparable to OpenAI-o1.
Distilling Small Models Surpassing OpenAI o1-mini
Alongside the open-source DeepSeek-R1-Zero and DeepSeek-R1, we distilled 6 small models from the outputs of DeepSeek-R1 and open-sourced them to the community, among which the 32B and 70B models achieved results that benchmark against OpenAI o1-mini in multiple capabilities.
HuggingFace Link: https://huggingface.co/deepseek-ai
1. Open your browser and enter the vscode download link: https://code.visualstudio.com/
2. After downloading, click the software and proceed with the installation by clicking next.
3. Open VSCode and open a folder using VSCode.
2. Configure the RooCline Plugin.
Open the VSCode software and search for the RooCline plugin as shown below:
After clicking the install button, the small rocket icon will appear.
3. Register for DeepSeek-R1 and Create API_Key
1. Open your browser and enter the address: https://platform.deepseek.com/sign_in, enter your phone number and verification code to register.
4. Configure DeepSeek-R1’s API_KEY
Copy the API_KEY created above into the plugin, the specific steps are as follows:
1. Select the robot icon -> 2. Select the gear icon -> 3. Enter the DeepSeek Api Key and select the model: DeepSeek-Reasoner -> 4. Click the Done button.
Click the share button to move the window to the right side for display
Click the <1> share button to move the window to the right side for display
Check the “Auto-approve” option to avoid manual intervention, allowing for free coding.
5. Start Writing the 2025 Spring Festival Couplets!
Input prompt: You are a programming expert, it’s 2025, please help me write a couplet that starts with a snake, the writing should contain joy and look forward to a big effort in 2025, and it should be implemented using JavaScript and HTML, with a fireworks effect animation.
Started writing automatically
It was completed in less than three seconds, let’s take a look at the effect:
Welcome to like, share, and collect!!!
