ChinaFeaturedtop-pickschina

DeepSeek

Strong coding and reasoning, cost-effective, good Chinese support.

getModelKey links official API key pages only. We do not sell, generate, or broker API keys.

Overview

DeepSeek provides coding- and reasoning-strong models through its platform API. Its current V4 endpoints are deepseek-v4-flash and deepseek-v4-pro, both with thinking and non-thinking modes.

Developers often pick DeepSeek for cost-effective Chinese-friendly coding agents and long reasoning traces without hosting the weights themselves.

How to Get an API Key

Register or sign in at platform.deepseek.com.
Open API Keys at https://platform.deepseek.com/api_keys and create a new key.
Copy the key when it is displayed and keep it out of public repos.
Call DeepSeek’s OpenAI-compatible chat APIs with that key using platform.deepseek.com API docs.

DeepSeek V4 models

Two depths of V4 intelligence

Choose Flash for throughput and API flexibility, or Pro for the higher-capability V4 tier.

CNY per 1M tokens

DeepSeek V4 Flash

High-throughput V4 model with the broadest API-format support.

Model ID: deepseek-v4-flash
Version: DeepSeek-V4-Flash-0731
Thinking: Non-thinking + thinking (default)

Cached input: ¥0.02
Uncached input: ¥1.00
Output: ¥2.00

Context: 1M
Max output: 384K
Concurrency: 2500
Responses API: Supported
API formats: OpenAI + Anthropic compatible
Features: JSON Output, Tool Calls, Anthropic API, prefix completion, FIM (non-thinking only)

DeepSeek V4 Pro

Higher-priced V4 tier for demanding reasoning and production workloads.

Model ID: deepseek-v4-pro
Version: DeepSeek-V4-Pro
Thinking: Non-thinking + thinking (default)

Cached input: ¥0.025
Uncached input: ¥3.00
Output: ¥6.00

Context: 1M
Max output: 384K
Concurrency: 500
Responses API: Not yet supported
API formats: OpenAI + Anthropic compatible
Features: JSON Output, Tool Calls, Anthropic API, prefix completion, FIM (non-thinking only)

DeepSeek currently warns that API prices are expected to increase significantly. Verify before purchase or deployment.

Official DeepSeek price snapshot checked August 10, 2026.Explore the complete DeepSeek pricing guide →

Common Models

DeepSeek V4 FlashDeepSeek V4 ProDeepSeek-V3.2DeepSeek-Coder-V3DeepSeek-R1DeepSeek-Reasoner

FAQ

Are DeepSeek APIs OpenAI-compatible?

Yes. The platform documents OpenAI-compatible endpoints; use your DeepSeek key and the official base URL.

Which DeepSeek models need a platform key?

Current V4 cloud endpoints use deepseek-v4-flash and deepseek-v4-pro. Older model lines may remain available; confirm current IDs in the official docs and console.

Are current DeepSeek V4 prices permanent?

No guarantee. DeepSeek currently warns that it plans a significant API price increase. Treat our figures as a dated snapshot and verify before recharging.

Pillar

Complete guide to China's major AI API platforms. Compare DeepSeek, Kimi/Moonshot, Qwen, ERNIE, and Doubao on pricing, registration, payment methods, and how to choose the right one for your project.

DeepSeek API Key Complete Guide (2026): Setup, Pricing & Usage

Pillar

Complete guide to DeepSeek API: how to get your key, understand pricing, use OpenAI-compatible SDK, and leverage DeepSeek-V3 and DeepSeek-R1 models effectively.

AI API Token Price Comparison 2026: GPT-5.6, Claude Fable 5, Gemini 3.6, DeepSeek & More

A practical breakdown of input and output token prices per 1M tokens across the leading AI providers in 2026, with guidance on choosing the right model for your budget and use case.

DeepSeek V4 Flash vs V4 Pro: Models & API Pricing

Compare official DeepSeek V4 Flash and V4 Pro pricing, cache rates, 1M context, 384K output, API compatibility, features, concurrency limits, and the announced price increase.

Free AI API Keys: Complete List of Free Tier Providers (2026)

Several AI providers offer free tiers or credits. Here's a complete comparison of which platforms give free access, how much you get, and the limits you should know about.

How to Get a DeepSeek API Key (2026 Step-by-Step Guide)