Google Launches New AI Models & Gemini Updates - Latest News

Google Unveils Gemini 2.0 Pro Experimental AI Model

On Wednesday, Google officially launched its highly anticipated Gemini 2.0 Pro Experimental, its newest flagship AI model. This release occurred alongside announcements of other advancements in their AI offerings.

The company is also integrating its 'reasoning' model, Gemini 2.0 Flash Thinking, directly into the Gemini app, expanding its accessibility to users.

Competition and Context

The release of these models arrives as the technology sector closely observes the emergence of more affordable AI reasoning models from DeepSeek, a Chinese AI startup.

DeepSeek’s models have demonstrated performance levels that either match or exceed those of leading AI models developed by prominent American technology companies.

Furthermore, businesses can gain access to DeepSeek’s capabilities through their API at a significantly lower cost.

Both Google and DeepSeek introduced AI reasoning models in December; however, DeepSeek’s R1 garnered considerable attention. Google’s move with Gemini 2.0 Flash Thinking may be a strategy to increase visibility for its own model within its established user base.

Gemini 2.0 Pro: A Leading Model

Gemini 2.0 Pro, the successor to the Gemini 1.5 Pro model launched last February, is now positioned by Google as the leading model within its Gemini AI family.

An initial announcement regarding Gemini 2.0 Pro’s release appeared in the Gemini app’s changelog approximately a week prior, but this week’s launch is the official rollout.

An experimental version of the model is now available on Wednesday through Google’s AI development platforms, Vertex AI and Google AI Studio. It will also be accessible to subscribers of Gemini Advanced via the Gemini app.

Enhanced Capabilities

According to Google, the new Gemini Pro model demonstrates exceptional performance in coding tasks and excels at processing intricate prompts.

It also exhibits “better understanding and reasoning of world knowledge” compared to Google’s previous models.

Gemini 2.0 Pro is capable of utilizing tools such as Google Search and executing code on behalf of users, enhancing its functionality.

Impressive Context Window

The model boasts a 2 million token context window, allowing it to process approximately 1.5 million words in a single instance.

To illustrate, Google’s latest AI model could accommodate all seven books of the Harry Potter series within a single prompt, with approximately 400,000 words of capacity remaining.

Gemini 2.0 Flash and Flash-Lite

Google is also making Gemini 2.0 Flash generally available on Wednesday. This model, initially announced in December, is now accessible to all Gemini app users.

Finally, in a move potentially designed to counter the excitement surrounding DeepSeek’s models, Google is introducing Gemini 2.0 Flash-Lite, a new, more cost-effective AI model.

The company claims this model surpasses the performance of its Gemini 1.5 Flash model while maintaining the same price point and speed.

TechCrunch has an AI-focused newsletter! Sign up here to get it in your inbox every Wednesday.