Revolutionizing AI Economics
Chinese startup **DeepSeek** has announced a milestone in artificial intelligence economics, claiming it trained its advanced **AI model** for just **$294,000** as of September 2025. This cost, a fraction of traditional industry benchmarks, signals major progress in large language model (LLM) accessibility and efficiency[1].DeepSeek vs. ChatGPT-4: Cost and Features
AI model training and operational expenses have long deterred smaller entrants. DeepSeek's achievement stands out when compared to global leaders like ChatGPT-4, which has historically required tens or even hundreds of millions of dollars to develop large-scale models[1].- DeepSeek's foundational model training cost: approx. $5.6 million for V3 model, with latest advanced model trained for $294,000.
- OpenAI's ChatGPT-4: creation estimated at over $100 million, with high operational expenses for API access[1].
Operational Costs: API Pricing Breakdown
For businesses and developers, real-world costs center on API usage and fine-tuning rather than initial training investments[1]. DeepSeek’s economics further shine here:- deepseek-chat: $0.27 per million input tokens and $1.10 per million output tokens.
- deepseek-reasoner: $0.55 per million input tokens and $2.19 per million output tokens.
- Cache hit pricing for repeated inputs: as low as $0.07 (chat), $0.14 (reasoner).
- This is up to 200 times cheaper than some major competitors, including OpenAI APIs, for high-volume uses.
- $5.00 per million input tokens and $15.00 per million output tokens.
- More affordable option with GPT-4o-mini: $0.15 per million tokens input, $0.60 output.
- GPT-4o: $25.00 per million tokens for training.
- GPT-4o-mini: $3.00 per million tokens.
Comparative Features: DeepSeek vs. ChatGPT-4
| Feature | DeepSeek AI (V3/R1) | ChatGPT-4 (GPT-4o / 4o-mini) |
|---|
