nexusflow
Online

DeepSeek Series

by DeepSeek AI

DeepSeek is a leading open-source large model known for its exceptional cost-effectiveness. It stands out in code generation and math reasoning, making it an ideal choice for budget-sensitive scenarios. Text models can be accessed via three public protocols: OpenAI Chat, Anthropic Messages, and Responses API.

Protocols

Core Advantages

💰
Highly cost-effective
About 1/10 the price of peers
🔓
Open & transparent
Weights are public and auditable
💻
Coding powerhouse
Outstanding programming ability
🧮
Math reasoning
Strong complex computation

Available Models

DeepSeek V4 Pro

NEW
deepseek-v4-pro
V4FlagshipReasoning

DeepSeek V4 flagship reasoning model, ideal for complex math, long-horizon decision-making, and code agent tasks.

Context Window
1,000,000
Max Output
16,384
Input Price
$12/M
Output Price
$24/M
Complex reasoningCode agentFunction calling

DeepSeek V4 Flash

NEW
deepseek-v4-flash
V4Ultra-fastHigh-concurrency

DeepSeek V4 high-speed version, low latency and high throughput, ideal for online Q&A, customer support, and high-concurrency gateway scenarios.

Context Window
1,000,000
Max Output
16,384
Input Price
$1/M
Output Price
$2/M
Low latencyHigh throughputOnline chat

DeepSeek R1

NEW
deepseek-r1
ReasoningChain-of-thought

Reasoning-enhanced model that analyzes complex problems via chain-of-thought. Excels at math, programming, and logical reasoning.

Context Window
64,000
Max Output
8,000
Input Price
$4/M
Output Price
$16/M
Chain-of-thoughtCode generationMath reasoning

DeepSeek V3

NEW
deepseek-v3
RecommendedCost-effective

Latest-generation general-purpose model with an MoE architecture, with all-round improvements in code, math, and general ability. Extremely cost-effective.

Context Window
64,000
Max Output
8,000
Input Price
$2/M
Output Price
$8/M
Code generationFunction callingMultilingual

Usage Example

from openai import OpenAI

client = OpenAI(
    api_key="sk-air-your-key",
    base_url="https://nexusflow.hk/v1",
)

# Use DeepSeek R1 for reasoning
response = client.chat.completions.create(
    model="deepseek-r1",
    messages=[
        {"role": "user", "content": "Reason step by step: if A > B and B > C, what is the relationship between A and C?"}
    ],
    max_tokens=2048,
)

# The R1 model shows its chain-of-thought process
print(response.choices[0].message.content)

Related Docs