Wed Feb 05 17:14:19 UTC 2025: ## Google Unveils Cheaper Gemini AI Models to Compete with DeepSeek

**Mountain View, CA –** Google announced significant updates to its Gemini family of large language models on Wednesday, introducing a new line of more affordable options designed to compete directly with lower-cost rivals like China’s DeepSeek. The move comes amidst investor scrutiny of AI development costs following DeepSeek’s revelation of surprisingly low training expenses.

The highlight of the announcement is Gemini 2.0 Flash-Lite, an even cheaper variant of Google’s existing Gemini Flash model. Gemini 2.0 Flash, previously available in a developer preview, is now publicly released, while Gemini Pro 2.0 enters testing phases. Google’s CTO of DeepMind, Koray Kavukcuoglu, cited positive feedback on Flash 1.5 as the impetus for developing Flash-Lite. While Flash-Lite offers significant cost savings – processing 1 million tokens costs just $0.019 – it’s worth noting that this is still more expensive than DeepSeek’s current offering ($0.014, set to increase fivefold on February 8th), but significantly undercuts OpenAI’s cost-efficient model ($0.075).

The pricing strategy reflects growing pressure on AI companies to demonstrate cost efficiency. Recent investor concerns, fueled by DeepSeek’s low training costs, have impacted Alphabet’s stock price. Although Alphabet, Microsoft, and Meta all continue to invest heavily in AI, the focus on cost-effectiveness is becoming increasingly prominent. Google’s new pricing represents a direct response to this market trend and a clear attempt to maintain competitiveness in the rapidly evolving landscape of AI technology.

Read More