Skip to content

DeepSeek R1 + CherryStudio - Ultimate Solution for Chinese AI, Local Deployment Can Match ChatGPT!

CherryStudio

I. Background: Why Do We Need Domestic AI?

As a heavy user of AI technology, I used to be a loyal fan of ChatGPT. However, with increased usage frequency, I gradually realized the risks of relying on overseas AI services:

  1. Account Banned: One late night, my ChatGPT account was suddenly banned for unknown reasons.
  2. Insufficient Chinese Support: When searching Chinese content on Google, frequent "request rejected" prompts appeared.
  3. Privacy Issues: Sensitive data uploaded to overseas servers poses leakage risks.

These issues made me realize we need a top-tier AI model truly belonging to the Chinese world. Thus, I discovered this domestic CP combination: DeepSeek R1 and CherryStudio.

II. DeepSeek R1: New Benchmark for Chinese AI

🔥 Core Advantages

  1. Powerful Chinese Understanding Capability
  2. Supports understanding of idioms, slang, internet memes, even dialects.

  3. Example: Input "内卷" (involution) and "躺平" (lying flat), the model can accurately explain their social background and cultural meaning.

  4. Multi-modal Input Support

  5. Supports PDF, Word, PPT and other document analysis, easily extracting key information.

  6. Actual test: Upload a 50-page academic paper, the model can generate summary and key conclusions within 1 minute.

  7. Knowledge Base Enhancement

  8. Users can upload custom knowledge bases to improve model performance in specific domains.
  9. Example: After lawyers upload legal statutes, the model can accurately cite relevant regulations to answer questions.

💡 Performance Comparison with GPT-4

Feature DeepSeek R1 GPT-4
Chinese Understanding ✅ Idioms, slang, internet memes ❌ Awkward literal translation
Response Speed ⏱️ Local version second-level response ⌛ Need to queue for API
Hardware Threshold 💻 Laptop can run 🚀 Depends on top graphics cards
Privacy Protection 🔐 Supports local deployment 🌐 Data uploaded to cloud

III. CherryStudio: AI Interaction Artifact

🛠️ Core Functions

  1. Multi-Model Dialogue
  2. Dialogue with multiple AI models simultaneously, comparing their answers.

  3. Example: Let DeepSeek R1 and ChatGPT answer a question simultaneously, choose the best answer.

  4. Knowledge Base Management

  5. Supports importing PDF, Word, PPT, web pages and other data sources.

  6. Data undergoes cleaning, segmentation, and vectorization processing, stored in vector database.

  7. User-Friendly Interface

  8. Provides graphical interface, simplifying model selection, API key management, and knowledge base configuration.

🚀 Usage Scenarios

  • Academic Research: Upload paper PDF, quickly generate summary and key conclusions.
  • Enterprise Knowledge Management: Build industry-specific knowledge bases, improve employee efficiency.
  • Personal Assistant: Local deployment, provides intelligent services while protecting privacy.

IV. Detailed Tutorial: Playing with DeepSeek R1 from Scratch

🛠️ Preparation Tools

🚀 Deployment Steps

  1. Cloud Experience Version (Suitable for Beginners)
  2. Register on SiliconFlow Platform, get API key.
  3. Open CherryStudio, input API key, select DeepSeek R1 model.

  4. Start conversation, enjoy the powerful capability of 671B parameter model.

  5. Local Hardcore Version (Suitable for Tech Enthusiasts)

  6. Install Ollama: bash # macOS/Linux curl -fsSL https://ollama.ai/install.sh | sh # Windows winget install ollama
  7. Download DeepSeek R1 model: bash ollama run deepseek-r1-32b-chinese

  8. Configure CherryStudio:

    • Open CherryStudio, select "Local Model", input Ollama's API address.
    • Select downloaded DeepSeek R1 model, start local conversation.

  9. Knowledge Base Loading (Advanced Feature)

  10. Open CherryStudio's knowledge base management interface.
  11. Drag PDF/Word/PPT files to interface, wait for system to complete vectorization processing.
  12. Input questions, the model will generate more accurate answers combining knowledge base content.

V. Why is DeepSeek R1 the Future of Chinese AI?

  1. Cultural Adaptability
  2. Understands idioms, slang, and internet memes in Chinese context.

  3. Example: Input "内卷" (involution) and "摸鱼" (slacking), the model can accurately explain their social background.

  4. Data Sovereignty

  5. Supports local deployment, sensitive data doesn't need to be uploaded to cloud.

  6. Example: Enterprises can deploy on internal servers, ensuring data security.

  7. Continuous Evolution

  8. Achieves model "lifelong learning" through knowledge bases.
  9. Example: Upload latest industry reports, the model can immediately grasp latest trends.