DeepSeek

When to Use

• Cost Efficiency: The API is incredibly cheap.
• Reasoning: DeepSeek-R1 uses Chain-of-Thought reinforcement learning (like OpenAI o1) but is open weights.
• Coding: DeepSeek-Coder-V2 is a top-tier coding model.

Core Concepts

MLA (Multi-Head Latent Attention)

DeepSeek-R1

Best Practices (2025)

• Use R1 for Math/Logic: It rivals o1-preview in math benchmarks.
• Local Distillations: Run

  DeepSeek-R1-Distill-Llama-70B

  locally for private reasoning.

• Don't suppress thoughts: When using R1, the "thought" trace is valuable for debugging the model's logic.

References

• DeepSeek GitHub