Soon after the Gemini 3 Pro launch, Google, on 17 December, released the Gemini 3 Flash AI model, which has also become the default AI model on the Gemini app. The tech giant allows free users to ask limited queries to its Thinking model, while the rest of the queries are directed to the non-reasoning model, which is denoted as ‘Fast’ in the Gemini app.
“Gemini 3 Flash demonstrates that speed and scale don’t have to come at the cost of intelligence. It delivers frontier performance on PhD-level reasoning and knowledge benchmarks like GPQA Diamond (90.4%) and Humanity’s Last Exam (33.7% without tools), rivaling larger frontier models, and significantly outperforming even the best 2.5 model, Gemini 2.5 Pro, across a number of benchmarks,” Google said in a blog post.
Gemini 3: What’s new?
Google says Gemini 3 Flash is three times faster than the Gemini 2.5 Pro model on the Artificial Analysis benchmark, while running at a fraction of the cost. For context, Gemini 2.5 Pro had remained among the top-ranking models across benchmarks for months until Google released its Gemini 3 Pro models last month.
In the coveted Humanity’s Last Exam benchmark, Gemini 3 Flash gets a score of 33%, which is higher than the scores of Grok 4.1 Fast (17.6%) and Claude Sonnet 4.5 (13.7%), while being just below GPT-5.2 (34.5%).
In SWE-bench Verified, a benchmark for coding-related tasks, Gemini 3 Flash got a score of 78%, placing it ahead of Claude 4.5 Sonnet (77.2%) and closely behind GPT-5.2 (80.0%).
Meanwhile, in the ARC-AGI-2 benchmark for visual reasoning-related tasks, Gemini 3 Flash outperformed not just Claude and Gemini 2.5 Pro, but even the latest Gemini 3 Pro model.
“When processing at the highest thinking level, Gemini 3 Flash is able to modulate how much it thinks. It may think longer for more complex use cases, but it also uses 30% fewer tokens on average than 2.5 Pro, as measured on typical traffic, to accurately complete everyday tasks with higher performance,” Google explained.
Where can you use Gemini 3 Flash?
Gemini 3 Flash is now already the default model in Gemini, and users don’t need to pay any extra cost to leverage its capabilities. Meanwhile, the model also powers AI Mode in Search, which allows users to ask questions in a conversational style, much like Perplexity.
Google says Gemini 3’s multimodal reasoning capabilities can help users “see, hear, and understand” any type of information faster. The company adds that users can ask Gemini to understand their videos and images, and even turn that information into an action plan within seconds.
About the Author
