Google rolls out cheaper AI model as industry scrutinizes costs

Google has announced several updates to its Gemini portfolio, including a budget-friendly product, amid growing demand for low-cost AI models driven by the rise of Chinese competitor DeepSeek.

“We’re releasing a new model, Gemini 2.0 Flash-Lite, our most cost-efficient model yet, in public preview in Google AI Studio and Vertex AI,” the company said in a blog post.

The updated lineup now includes several Gemini variants tailored to different price points and performance levels.

In addition to Flash-Lite, Google has made Gemini 2.0 Flash generally available after a developer preview in December and has begun testing an updated version of its flagship “Pro” model.

The updated Gemini 2.0 Flash would be available via the Gemini API in Google AI Studio and Vertex AI, allowing developers to build production applications with 2.0 Flash.“The Flash series of models is popular with developers as a powerful workhorse model, optimal for high-volume, high-frequency tasks at scale and highly capable of multimodal reasoning across vast amounts of information with a context window of 1 million tokens,” Google said. “2.0 Flash is now generally available to more people across our AI products, alongside improved performance in key benchmarks, with image generation and text-to-speech coming soon.”

Flash-Lite 2 features a 1-million token context window and supports multimodal input, similar to Flash 2, Google added. It can generate captions for around 40,000 unique images at a cost of less than a dollar on Google AI Studio’s paid tier.

DeepSeek’s disruption in AI

The rising costs of AI development have drawn heightened investor scrutiny after Chinese AI firm DeepSeek revealed it spent less than $6 million on the final training run of one of its models – far below what developers at leading US AI firms estimate for similar systems.

DeepSeek’s disclosure has fueled investor concerns, surfacing in earnings calls at Alphabet, Microsoft, and Meta, as stakeholders question the long-term sustainability of soaring AI investments.

“The introduction of DeepSeek models has disrupted the generative AI landscape through several key developments, including cost-effective model development, the democratization of AI access, and geopolitical and security considerations,” said Kartikey Kaushal, senior analyst at Everest Group. “The impact of these developments was felt globally, as major firms like Nvidia, Microsoft, and Alphabet saw declines in their stock prices.”

Competition in the AI sector intensified further over the Lunar New Year holidays last month as Alibaba Group unveiled an upgraded AI model, claiming it outperforms DeepSeek. The move raises the stakes in the battle for low-cost, high-performance AI solutions.

“From a vendor standpoint, there will be increased pressure to deliver a broader range of affordable AI solutions,” Kaushal said. “Vendors like OpenAI and Microsoft will need to reassess their pricing models to stay competitive and attract investments. Additionally, they must balance R&D spending with revenue generation.”

Despite the cost concerns, Google, Microsoft, and Meta have signaled plans to maintain substantial capital spending in the field.

Implications for enterprises

Google’s launch of the Flash-Lite model marks a direct response to its Chinese rival but also signals a broader shift in the enterprise AI landscape, where cost efficiency is becoming a key battleground.

For businesses integrating AI, this shift could lead to a reallocation of budgets, moving spending away from costly model training and development toward AI integration, automation, and workflow optimization. However, analysts caution that it may come with trade-offs.

“Enterprises opting for low-cost AI models must weigh trade-offs in performance, security, and data privacy,” Kaushal said. “With fewer parameters, these models may have weaker contextual understanding and reasoning, making them less effective for complex tasks.”

Beyond performance considerations, security risks include weaker defenses against adversarial attacks and data poisoning. Shared infrastructure also raises concerns over data privacy, potential leaks, and regulatory compliance challenges.

Hidden costs – such as additional validation layers, security measures, and reputational risks from performance issues – may further complicate decision-making, underscoring the need for a balanced approach.