Deepseek is a new AI model from China. It outperforms ChatGPT, Gemini, and Claude AI. Its app became the most downloaded free app in the US. As a consequence, Nvidia, Microsoft, and Meta shares dropped. It is important to understand why Deepseek is a game changer. Letās delve into this more.
What is Deepseek?
In 2023, Liang Wenfeng founded Deepseek, an AI research lab located in Hangzhou. Liang was the head of a hedge fund that relied on AI for financial analytics. He recruited a team of graduates from top-tier Chinese universities like Tsinghua and Peking. Deepseek’s models are open source and have a development cost of less than $6 million, in stark contrast to competitors such as OpenAI and Meta.
How Does Deepseek Compare to ChatGPT and Others?
Deepseekās app reached first place in downloads within the US, UK, and China. It provides unlimited free usage, unlike paid models such as ChatGPT-4 and Claude Sonnet. Google Gemini charges no fee but is based on older models and has usage limitations. The R1 model of Deepseek is said to think like a human. It articulates its reasoning before concluding with answers to questions. Tests demonstrate it is on par with the most recent OpenAI technology.
Why is Deepseek Challenging Nvidia?
Deepseek utilizes lower-end hardware, so unlike the majority of AI systems reliant on Nvidia chips, Deepseek can be run on cheaper hardware. This is a crucial asset, as the US is no longer allowing advanced chip exports to China. Deepseek is demonstrating to the world that AI can function more than optimally without top-tier hardware.
What Makes Deepseek Special?
- Open-Source: Programmers from around the world have the ability to utilize Deepseek-V3 without any costs being incurred.
- Affordable: Deepseek V3 was made for under $6 million as opposed to billions spent by competitors.
- Explanatory: It clearly defines its operating process.
- No Payment: Paid competitors have put a paywall behind their systems, whereas Deepseek has not.
How is Deepseek Changing the AI Industry?
The success of Deepseek shook the stock market. Nvidia, Meta, and Microsoft shares have since then continued to tumble. Deepseek is aiding or utilizing the idea that the AI market can be stable without considerable spending. The budget for AI can easily be challenged with the inclusion of cheap Chinese equipment, which poses a larger threat to the American economy as analysts have claimed. Deepseek works wonders in proving how open-source AI might very well give birth to new innovations.
Challenges for Deepseek
- Trust: There is still a requirement for users to see and believe in the reliability and safety offered.
- Competition: Remaining rivals Deep AII and Google have compiled a hefty amount of funds.
- Regulations: Government actions could hinder AI by implementing tougher regulation policies.
The Future of Deepseek
The primary model Duke seeks to bridge the gap between culture and language when it comes to state-of-the-art AI innovation. Its affordable model will attract startups from around the globe. It also anticipates that the big tech industry will have to rethink their pricing strategy with its expansion.
DeepSeek-V3 Capabilities
| Benchmark (Metric) | DeepSeek V3 | DeepSeek V2.5 | Qwen2.5 | Llama3.1 | Claude-3.5 | GPT-4o | |
|---|---|---|---|---|---|---|---|
| 905 | 72B-Inst | 405B-Inst | Sonnet-1022 | 513 | |||
| Architecture | MoE | MoE | Dense | Dense | – | – | |
| # Activated Params | 37B | 21B | 72B | 405B | – | – | |
| # Total Params | 671B | 236B | 72B | 405B | – | – | |
| English | MMLU (EM) | 88.5 | 80.6 | 85.3 | 88.6 | 88.3 | 87.2 |
| MMLU-Redux (EM) | 89.1 | 80.3 | 85.6 | 86.2 | 88.9 | 88 | |
| MMLU-Pro (EM) | 75.9 | 66.2 | 71.6 | 73.3 | 78 | 72.6 | |
| DROP (3-shot F1) | 91.6 | 87.8 | 76.7 | 88.7 | 88.3 | 83.7 | |
| IF-Eval (Prompt Strict) | 86.1 | 80.6 | 84.1 | 86 | 86.5 | 84.3 | |
| GPQA-Diamond (Pass@1) | 59.1 | 41.3 | 49 | 51.1 | 65 | 49.9 | |
| SimpleQA (Correct) | 24.9 | 10.2 | 9.1 | 17.1 | 28.4 | 38.2 | |
| FRAMES (Acc.) | 73.3 | 65.4 | 69.8 | 70 | 72.5 | 80.5 | |
| LongBench v2 (Acc.) | 48.7 | 35.4 | 39.4 | 36.1 | 41 | 48.1 | |
| Code | HumanEval-Mul (Pass@1) | 82.6 | 77.4 | 77.3 | 77.2 | 81.7 | 80.5 |
| LiveCodeBench (Pass@1-COT) | 40.5 | 29.2 | 31.1 | 28.4 | 36.3 | 33.4 | |
| LiveCodeBench (Pass@1) | 37.6 | 28.4 | 28.7 | 30.1 | 32.8 | 34.2 | |
| Codeforces (Percentile) | 51.6 | 35.6 | 24.8 | 25.3 | 20.3 | 23.6 | |
| SWE Verified (Resolved) | 42 | 22.6 | 23.8 | 24.5 | 50.8 | 38.8 | |
| Aider-Edit (Acc.) | 79.7 | 71.6 | 65.4 | 63.9 | 84.2 | 72.9 | |
| Aider-Polyglot (Acc.) | 49.6 | 18.2 | 7.6 | 5.8 | 45.3 | 16 | |
| Math | AIME 2024 (Pass@1) | 39.2 | 16.7 | 23.3 | 23.3 | 16 | 9.3 |
| MATH-500 (EM) | 90.2 | 74.7 | 80 | 73.8 | 78.3 | 74.6 | |
| CNMO 2024 (Pass@1) | 43.2 | 10.8 | 15.9 | 6.8 | 13.1 | 10.8 | |
| Chinese | CLUEWSC (EM) | 90.9 | 90.4 | 91.4 | 84.7 | 85.4 | 87.9 |
| C-Eval (EM) | 86.5 | 79.5 | 86.1 | 61.5 | 76.7 | 76 | |
| C-SimpleQA (Correct) | 64.1 | 54.1 | 48.4 | 50.4 | 51.3 | 59.3 |
Among the open-source models, DeepSeek-V3 scores the highest in terms of speed and performance. It is even competitive with Open AIās GPT-4o, being a closed-source proprietary model.
Deepseek is putting forward the proposition that advanced AI does not have to be expensive or proprietary. It goes head-to-head with the likes of ChatGPT and Nvidia while also shifting the patterns of global technology. The race for AI just got more intense.
