Google has slashed the price of its new AI Ultra plan by 60%, from $249.99 to $99.99 per month, to introduce its 'agentic' Gemini Spark assistant. This move comes even as its underlying models show only 68.8% accuracy on factual grounding tests, according to Mashable. The AI Ultra tier, now $99.99 per month, includes 20 TB storage and Gemini Spark beta access, as Findskill reports. Google is launching a premium 'agentic' AI assistant at a significantly reduced but still high price, but the foundational models it relies on still demonstrate notable accuracy limitations in critical areas like factual grounding. Therefore, the current AI Ultra offering appears to be a strategic move by Google to capture market share in the nascent agentic AI space, requiring users to weigh the cost against the promise of future advancements rather than current perfection.
The Current State of Gemini's Performance
Google's Gemini 3 Pro shows an overall accuracy of 68.8% on the FACTS Grounding test, while scoring 83.8% on the FACTS Search benchmark, both according to Mashable. This suggests the model is better at finding information than it is at verifying its factual accuracy, a crucial distinction for an "agentic" AI. Companies adopting Google's 'agentic' Gemini Spark are effectively signing up to be beta testers, trading potential future productivity for current reliability risks.
Google's Tiered Strategy and Premium Offerings
Google offers a range of AI tiers, from the AI Plus at $7.99 per month with 200 GB storage, to the AI Pro at $19.99 per month with 5 TB storage and Gemini 3.1 Pro with 1M context, as reported by Findskill. At the top end, the AI Ultra Premium tier costs $200 per month, including 20 TB storage with 20x Pro usage, making it Google's most expensive offering. This elaborate tiered structure aims to attract diverse users, but the Ultra and Ultra Premium tiers, along with Gemini Spark, clearly target early adopters willing to invest in cutting-edge AI. The significant 60% price cut for the AI Ultra plan suggests Google is aggressively pursuing market share in the 'agentic' AI space, prioritizing a premium vision over a fully perfected product.
Future Outlook and Safety Commitments
Google plans to release more information about the newest models' safety evaluations alongside the rest of the Gemini 3.5 model series, expected in June, Mashable reports. However, launching the AI Ultra plan before these full safety evaluations are published suggests Google prioritizes speed to market over comprehensive vetting, potentially exposing early adopters to unforeseen issues.
Frequently Asked Questions
What are agentic AI capabilities?
Agentic AI refers to systems designed to take actions and achieve goals autonomously, often by breaking down complex tasks into smaller steps. For example, Gemini Spark is built to act as a 24/7 assistant, handling tasks like scheduling or information retrieval without constant user input, as detailed by TechCrunch.
What kind of tasks can Gemini Spark perform?
Gemini Spark is designed for continuous background operation, managing tasks such as email triage, summarizing documents, or coordinating schedules. It aims to automate routine cognitive tasks, allowing users to offload repetitive work to the AI assistant, according to TechCrunch.
What are the risks of adopting agentic AI early?
Early adoption of agentic AI systems like Gemini Spark carries risks, including potential for factual inaccuracies due to foundational model limitations. Users may also encounter unexpected behavior or integration challenges as the technology is refined, requiring them to actively participate in identifying and correcting errors.
