DeepSeek
Chinese AI lab's open-weight reasoning model with frontier performance at extremely low cost
DeepSeek is a Chinese AI research lab whose R1 reasoning model shook the industry with frontier-level performance at a fraction of competitors' costs. The current V3.2 model unifies chat and reasoning into a single model at $0.28/$0.42 per MTok. R1 is free to use on their platform and available as open weights. DeepSeek V4, a multimodal trillion-parameter model, has been announced but not yet released as of March 2026.
Last updated: 2026-03-12 00:09:18
AI Models Powering DeepSeek
2 models from 1 vendor
DeepSeek V3.2
Primary Currentdeepseek-chat Extremely cost-efficient. Unified architecture supports both fast chat and deep reasoning via parameter toggle. 90% cache hit discount.
DeepSeek R1
Currentdeepseek-reasoner Extremely cost-efficient reasoning model. Explicit chain-of-thought. Open-weight (can self-host). Pricing unified with V3.2 — dramatically cheaper than original R1 pricing.
Model data sourced from official vendor documentation. View all models →
Recent Model Updates
DeepSeek Pricing
Free
API
Pay-per-use, significantly cheaper than competitors
Pricing verified as of research date. Contact vendor for current pricing.
DeepSeek Features
Code & Development
Conversation & Chat
Platform & Infrastructure
DeepSeek Pros & Cons
Pros
- + Free web interface with strong capabilities
- + Open-source models can be self-hosted
- + Extremely cheap API pricing — fraction of OpenAI costs
- + Strong reasoning and maths capabilities
Cons
- - Data privacy concerns — servers in China
- - English language quality behind Western competitors
- - Less polished interface and user experience
- - Smaller feature set than ChatGPT or Claude
DeepSeek Integrations
Company Info
- Company
- DeepSeek
- Founded
- 2023
- Headquarters
- Hangzhou, China
- Deployment
- Cloud
Get AI Tool Updates
Weekly reviews, pricing changes, and comparison guides.