DeepSeek R1 API Pricing

2/13/2025

Current Status of DeepSeek R1 API

DeepSeek R1 is an AI model developed in 2023 by the Hangzhou Research Laboratory in China, led by AI and finance expert Wenfeng Liang. The model's performance is comparable to mainstream models like ChatGPT, Gemini, and Claude. Currently, the official DeepSeek API has the following situation:

1. Service Status:

   • Official API is currently unstable with frequent unavailability
   • New user deposits suspended
   • Existing users face usage restrictions

2. Pricing Plan:

   • Input tokens: ¥1/million (cache hit) or ¥4/million (cache miss)
   • Output tokens: ¥16/million
   • ¥10 credit for first-time registration

3. Technical Features:

   • Supports 64K context length
   • Provides Chain of Thought reasoning capabilities
   • Uses smart caching mechanism, saving up to 90% cost on repeated queries

Due to the instability of the official API, various alternatives have emerged in the market, including third-party API services, local deployment solutions, and integrated services from major cloud providers. Here's a compilation of the main alternatives:

Service List and Price Comparison

Price Comparison Table

Note: Prices are per million tokens

International Service Providers

1. Fireworks AI

   • URL: https://fireworks.ai/models/fireworks/deepseek-r1
   • Features:
     • OpenAI format compatible
     • US deployment
     • $1 credit for first registration
     • DeepSeek V3 pricing: $0.9/1M

2. Openrouter

   • URL: https://openrouter.ai/deepseek/deepseek-r1:free
   • Model: DeepSeek R1 (free)
   • Launch date: January 20, 2024
   • Features:
     • Context length: 163,840 tokens
     • Input price: Free
     • Output price: Free
     • Performance comparable to OpenAI o1
     • Model size: 671B parameters (37B activated during inference)
   • Open source info:
     • Fully open-source model and technical report
     • MIT license: Free for commercial use and modification

3. Together AI

   • URL: https://api.together.ai/playground/chat/deepseek-ai/DeepSeek-R1
   • Features: Most stable third-party API, higher pricing

Local Deployment Solutions

1. Simple Installation

   • Ollama install: ollama run deepseek-r1:8b
   • AnythingLLM installation
   • lmstudio

2. Cloud Service Deployment

   • Microsoft Azure: https://ai.azure.com/
   • Amazon Bedrock: https://aws.amazon.com/cn/blogs/china/deepseek-r1-models-now-available-on-aws/

Chinese Service Providers

1. DeepSeek Official

   • URL: http://platform.deepseek.com
   • Free credit: ¥10 (≈$1.4)
   • Pricing:
     • Input tokens: ¥1 (≈$0.14)/million (cache hit) or ¥4 (≈$0.56)/million (cache miss)
     • Output tokens: ¥16 (≈$2.2)/million
   • Note: ByteDance, Alibaba Cloud, and paid Silicon Flow are currently stable

2. Silicon Flow

   • URL: https://cloud.siliconflow.cn/i/adxzw8w3af
   • Note: Free credits not applicable for full R1

3. Tencent Cloud

   • URL: https://cloud.tencent.com/act/pro/deepseek-api
   • Free: 500K tokens
   • API URL: https://api.lkeap.cloud.tencent.com/v1
   • Feature: Fast response time

4. Baidu Cloud Qianfan

   • URL: https://console.bce.baidu.com/qianfan/modelcenter/model/buildIn/list

5. ByteDance Volcano Engine

   • URL: https://console.volcengine.com/ark/region:ark+cn-beijing/experience
   • Free: 500K tokens

6. Alibaba Cloud

   • Free: 1M tokens

Other Platform Support

1. POE

   • URL: https://poe.com/DeepSeek-R1

2. Cursor

   • URL: https://cursor.sh

3. Monica

   • URL: https://monica.im

4. Lambda

   • URL: https://lambdalabs.com

5. Cerebras

   • URL: https://cerebras.ai

6. MixtureGPT

   • URL: https://mixturegpt.com

7. Perplexity

   • URL: https://perplexity.ai

8. China Supercomputing Internet

   • URL: https://chat.scnet.cn/#/home

9. 360 Nano AI Search

   • URL: https://n.cn

10. Metaso Search

    • URL: https://metaso.cn

11. Hix AI

    • URL: https://chat.hix.ai/deepseek-r1

Open Source Solution

Reverse-engineered API based on DeepSeek web version: https://github.com/LLM-Red-Team/deepseek-free-api

References

• https://meetcody.ai/blog/deepseek-r1-api-pricing/
• https://apidog.com/blog/deepseek-r1-review-api/