Google Unveils Affordable AI Models Amid Rising Cost Worries
Google Introduces Updates to Gemini Language Models
Alphabet’s Google recently unveiled updates to its Gemini family of large language models, including a new product line that aims to compete with low-cost artificial intelligence models, such as those offered by Chinese competitor DeepSeek.
New Models and Features
The tech giant offers a range of Gemini models that vary in price and performance. In addition to the existing “Flash” variant, Google has introduced a new, even more affordable model called “Flash-Lite.”
Google officially released Gemini 2.0 Flash to the general public after providing a preview to developers in December. Alongside this release, the company also introduced Flash-Lite and initiated testing for an updated version of its flagship “Pro” model.
According to Koray Kavukcuoglu, chief technology officer of Google’s DeepMind AI lab, the development of Flash-Lite was prompted by positive feedback on the 1.5 version of Flash. While Flash-Lite comes at a lower cost, Gemini 2.0 Flash is priced higher compared to its predecessor.
Cost Considerations and Investor Scrutiny
The cost associated with developing AI models and utilizing them has garnered increased attention from investors in recent weeks. This scrutiny intensified after DeepSeek disclosed that it spent less than $6 million on the final training run of a model. However, industry experts believe that the total cost for leading US AI firms is significantly higher than the figures reported by DeepSeek.
The rise of DeepSeek has raised questions during earnings calls for Alphabet, Microsoft, and Meta. Despite this, all three companies have indicated their commitment to substantial capital expenditures in the AI sector.
Market Impact and Pricing
Alphabet’s shares experienced a decline on Tuesday, partially attributed to investor concerns surrounding a planned capital expenditure increase that exceeded Wall Street’s expectations by 29%.
In terms of pricing, certain inputs for Gemini Flash-Lite are priced at $0.019 per 1 million tokens, which are units of data processed by an AI model. This pricing compares favorably to OpenAI’s flagship model, which costs $0.075 per token, and DeepSeek’s affordable model, priced at $0.014 per token. However, DeepSeek has announced an upcoming fivefold price increase on February 8th.
Overall, Google’s latest updates to its Gemini language models demonstrate the company’s commitment to innovation and competitiveness in the rapidly evolving field of artificial intelligence.