deepseek

DeepSeek

DeepSeek provides state-of-the-art open foundation models, including DeepSeek-V3 (Mixture-of-Experts) and DeepSeek-R1 for chain-of-thought mathematical reasoning and agentic workflows.

When to Use

• High-Performance Open-Weights Reasoning & Coding: DeepSeek-R1 and DeepSeek-V3 for mathematical reasoning, agentic logic, and code generation.
• Self-Hosted & Private LLM Deployments: Running enterprise reasoning models locally using vLLM, SGLang, or Ollama.
• Cost-Efficient API Scaling: Accessing state-of-the-art reasoning at a fraction of proprietary API costs.
• Chain-of-Thought (CoT) Verification: Parsing detailed <think> reasoning traces for auditability and verification.

Quick Start

from openai import OpenAI

# DeepSeek exposes an OpenAI-compatible API endpoint
client = OpenAI(
    api_key="your-deepseek-api-key",
    base_url="https://api.deepseek.com"
)

response = client.chat.completions.create(
    model="deepseek-reasoner", # DeepSeek-R1
    messages=[
        {"role": "user", "content": "Solve: How many r's are in strawberry? Think step by step."}
    ]
)

# Reasoning output is provided in reasoning_content
print("Thinking Process:\n", response.choices[0].message.reasoning_content)
print("Final Answer:\n", response.choices[0].message.content)

Core Concepts

Consuming DeepSeek API with OpenAI SDK Compatibility

Querying DeepSeek models with reasoning token handling:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ.get("DEEPSEEK_API_KEY"),
    base_url="https://api.deepseek.com"
)

response = client.chat.completions.create(
    model="deepseek-reasoner", # DeepSeek-R1 reasoning model
    messages=[
        {"role": "system", "content": "You are a quantitative systems engineer."},
        {"role": "user", "content": "Design an optimal concurrent lock-free queue algorithm in C++."}
    ],
    max_tokens=4096,
    temperature=0.6,
)

# DeepSeek-R1 returns reasoning content separately
reasoning = getattr(response.choices[0].message, 'reasoning_content', None)
final_answer = response.choices[0].message.content

if reasoning:
    print(f"=== Thinking Process ({len(reasoning)} chars) ===")
    print(reasoning[:500] + "...")

print("\n=== Final Response ===")
print(final_answer)