Overview
DeepSeek-R1 is a reasoning-focused model from the Chinese AI lab DeepSeek. It uses chain-of-thought techniques to excel at math, logic, and complex problem-solving. Available on Ollama for local deployment.
Sizes
1.5B / 7B / 8B / 14B / 32B / 70B
Context
64K-128K tokens
Specialty
Math & Logic
License
MIT
✅Strengths
- ✓Best-in-class for math and reasoning tasks
- ✓Chain-of-thought reasoning built-in
- ✓Competitive with o1 on math benchmarks
- ✓Multiple sizes for different hardware
- ✓Permissive MIT license
⚠️Weaknesses
- ✗Slower inference (reasoning overhead)
- ✗Overkill for simple tasks
- ✗Less optimized for creative writing
- ✗Requires local hardware
Best Use Cases
🧮 Math Problems
STEM, calculations
🧩 Logic Puzzles
Reasoning tasks
💻 Code Logic
Algorithm design
📊 Data Analysis
Complex reasoning
🔬 Research
Scientific problems
🎓 Education
Math tutoring
Ollama Command
ollama run deepseek-r1:14b
Available variants: deepseek-r1:1.5b, deepseek-r1:7b, deepseek-r1:8b, deepseek-r1:14b, deepseek-r1:32b, deepseek-r1:70b