AI War Heats Up: Google's Gemini 2.0 Flash Goes Public After DeepSeek Disruption – Key Details Inside

Big tech giant Google on Wednesday opened Gemini 2.0 Flash to the public Wednesday, allowing everyone, not just testers and developers, to access what the company touted as its "most capable" AI model suite so far.

Gemini 2.0 Flash was made available for tech developers and trusted testers in December, but this time, Google said the suite has been made "generally available" via the Gemini API in Google AI Studio and Vertex AI.

Google Hails Gemini Suite

On Wednesday, Google said it continues to invest in robust measures at it makes the Gemini AI model suite even more capable than it already is.

"Our Gemini 2.0 lineup was built with new reinforcement learnings techniques that use Gemini itself to critique its responses. This resulted in more accurate and targeted feedback and improved the model's ability to handle sensitive prompts, in turn," Koray Kavukcuoglu, CTO, Google DeepMind, on behalf of the Gemini team, said in a blog post.

What is Gemini 2.0?

Google introduced Gemini 2.0 in December as part of its efforts to jump into the agentic AI era. The AI model suite is expected to enhance generative AI further with multimodal capabilities and other cutting-edge AI features.

Key features of 2.0 Flash:

• Developer-friendly: Touted by developers as a "powerful workhorse model," as per Google, 2.0 Flash is perfect for high-volume, high-frequency tasks at scale.
• Multimodal reasoning: The model can process vast amounts of information with a context window of one million tokens, which are individual units of data that an AI model processes.
• Specialty in various tasks: 2.0 Flash is an excellent tool for summarization, image and video captioning, chat applications, and extraction of data from long documents.

The Gemini 2.0 family suite also includes the following, which aren't available for all as of now.

Gemini 2.0 Flash-Lite:

• 1.5 Flash outperformer: Available in Google AI Studio and Vertex AI in public preview, the 2.0 Flash-Lite model also has a one million token context window and outperforms 1.5 Flash "on the majority of benchmarks."
• Cost-efficient: Google said the 1.5 Flash-Lite is its most cost-efficient model so far but has the same speed as its predecessor.
• One-liner: The model can produce a relevant one-liner caption for around up to 40,000 unique photos at less than a dollar in the Google AI Studio paid tier.

Gemini 2.0 Pro Experimental:

• Solid coding: Available for now as an experimental model to developers in Google AI Studio and Vertex AI and to Gemini Advanced users, Google calls the model its "best model yet for coding performance and complex prompts."
• Better reasoning: Gemini 2.0 Pro can handle more complex prompts and has displayed better understanding and reasoning of word knowledge than any other AI model Google has released so far.
• Massive context window: With a context window at two million tokens, Gemini 2.0 Pro can process and analyze comprehensive data and can even call tools like Google Search.

To compare, Gemini 2.0 Flash-Lite's reasoning is at 51.5%, Gemini 2.0 Flash achieved 60.1%, while the experimental Gemini 2.0 Pro logged a reasoning accuracy of 64.7%.

Google Moves Following DeepSeek Shocker

Google's latest move came in the aftermath of the tech market shakeup delivered by DeepSeek, a Chinese startup's AI assistant that's supposedly been developed with a $6 million budget – an amount significantly lower than the funding budget that other U.S. tech giants used up for their AI projects, including OpenAI's ChatGPT.

Sundar Pichai, the CEO of Google parent company Alphabet, praised DeepSeek earlier this week, saying the startup's "tremendous" work on the project, also noting the Chinese AI model's launch presents an opportunity.

"Part of the reason we are so excited about the AI opportunity is we know we can drive extraordinary use cases because the cost of actually using it is going to keep coming down, which will make more use cases feasible," he said.