Google has officially launched the Gemini 3 Flash model, positioned as a cost-effective alternative designed to outperform its predecessor and rivals like OpenAI’s offerings. This new model will serve as the default configuration within the Gemini app and its AI search mode, marking a significant update just a month after the introduction of Gemini 3.
The Gemini 3 Flash model arrives six months following the rollout of the Gemini 2.5 Flash, showcasing impressive advancements in performance metrics. In benchmark tests, Gemini 3 Flash yields notable results, achieving a 33.7% score on Humanity’s Last Exam—a test designed to assess proficiency across various subjects. This score places it in close competition with top models; for context, Gemini 3 Pro scored 37.5%, while both Gemini 2.5 Flash and GPT-5.2 garnered scores of 11% and 34.5%, respectively.
Further demonstrating its capabilities, Gemini 3 Flash excelled in multimodality and reasoning tests, securing an impressive 81.2% on the MMMU-Pro benchmark, thus outpacing all competitors.
Global Rollout and Features
Beginning its global rollout, Google is replacing Gemini 2.5 Flash with Gemini 3 Flash in the Gemini app. Users seeking advanced functionalities such as math and coding can still opt for the Pro model. The new Flash model is particularly adept at analyzing multimodal content, enabling users to upload videos, sketches, or audio files and receive tailored insights or suggestions. Furthermore, it enhances user-query comprehension and is capable of producing more visual responses, including images and tables.
Users can also leverage the Gemini app to design app prototypes through simple prompts, enriching the overall user experience.
Enterprise and Developer Access
Google has announced that enterprises including JetBrains, Figma, and Cursor are already integrating the Gemini 3 Flash model via Vertex AI and Gemini Enterprise. Developers can access the model in preview through the API and Google’s newly launched coding tool, Antigravity.
According to Google, the Gemini 3 Pro model surpassed the SWE-bench verified coding benchmark with a score of 78%, falling just short of GPT-5.2. It is optimized for tasks such as video analysis and visual Q&A, distinguished by its speed and efficiency that supports rapid workflows.
Pricing Structure
The pricing for the Gemini 3 model has been set at $0.50 per million input tokens and $3.00 per million output tokens, slightly higher than the previous Gemini 2.5 Flash model. Despite the cost increment, Google asserts that Gemini 3 Flash delivers better performance while being three times faster and consuming 30% fewer tokens for cognitive tasks.
Since its inception, Gemini 3 has processed over one trillion tokens daily, a metric highlighting its robust performance amidst heightened competition with OpenAI. Recently, OpenAI responded to a noted decline in ChatGPT’s traffic by deploying its own updates, including GPT-5.2. While Google refrained from directly addressing its rivalry with OpenAI, it emphasized the industry-wide momentum, with innovations pushing the boundaries of AI capabilities.
In conclusion, the launch of the Gemini 3 Flash model underscores Google’s commitment to evolving its AI landscape, promising enhanced features and performance that respond to user needs and sector demands.
