Google has officially introduced Gemini 3 Flash, a new lightweight AI model designed to deliver high performance at lower cost and faster speeds. Built on the Gemini 3 architecture released last month, the model is positioned as a direct response to growing competition from OpenAI.
Google is not treating this as a niche release. Gemini 3 Flash is now the default model inside the Gemini app and Google Search’s AI Mode, replacing Gemini 2.5 Flash globally.
This move signals Google’s intent to scale advanced AI access across consumer, developer, and enterprise use cases.
Strong Performance Gains Over Previous Flash Models
Gemini 3 Flash arrives just six months after the launch of Gemini 2.5 Flash, but the improvement is substantial. On multiple industry benchmarks, the new model shows major gains and, in some areas, performs on par with top-tier models such as Gemini 3 Pro and GPT-5.2.
On Humanity’s Last Exam, a benchmark designed to measure reasoning and expertise across diverse fields, Gemini 3 Flash achieved a 33.7 percent score without tool use. This places it well ahead of Gemini 2.5 Flash, which scored 11 percent, and close to Gemini 3 Pro at 37.5 percent and GPT-5.2 at 34.5 percent.
In multimodal reasoning, the model performed even more strongly. On the MMMU-Pro benchmark, which evaluates understanding across text, images, and complex reasoning tasks, Gemini 3 Flash recorded an 81.2 percent score, outperforming all competing models in the comparison.
Default Model for Consumers Worldwide
Google is rolling out Gemini 3 Flash as the standard experience for users worldwide through the Gemini app. While the Flash model will handle most everyday queries, users can still manually switch to Gemini 3 Pro for advanced tasks such as complex mathematics or coding.
According to Google, Gemini 3 Flash is particularly strong at interpreting multimodal inputs. Users can upload short videos for feedback, share sketches for recognition, submit audio files for analysis, or request the model to generate quizzes and insights from uploaded content.
The company also says the model does a better job of understanding user intent and delivering more visual responses, including structured elements like tables and images.
Expanding Access Across Search and Creative Tools
Alongside the Flash rollout, Google has expanded availability of Gemini 3 Pro in Search across the United States. More U.S. users also now have access to the company’s advanced image generation model, Nano Banana Pro, directly through search experiences.
Gemini 3 Flash can also be used to quickly build application prototypes within the Gemini app using natural language prompts, further blurring the line between experimentation and production.
Enterprise and Developer Adoption Is Already Underway
On the enterprise side, Google confirmed that companies such as JetBrains, Figma, Cursor, Harvey, and Latitude are already using Gemini 3 Flash in production environments. The model is available through Vertex AI and Gemini Enterprise, making it accessible for large-scale business workflows.
Developers can access Gemini 3 Flash as a preview through Google’s API and within Antigravity, the company’s new coding tool introduced last month.
Google also highlighted the strength of Gemini 3 Pro in coding tasks. The model achieved a 78 percent score on the SWE-bench verified benchmark, placing it just behind GPT-5.2. Because of its speed and efficiency, Google positions Gemini 3 Flash as ideal for video analysis, data extraction, visual question answering, and repetitive workflows.
Pricing Designed for High-Volume Use
Gemini 3 Flash is priced at $0.50 per million input tokens and $3.00 per million output tokens. While this is slightly higher than Gemini 2.5 Flash, Google says the new model delivers significantly more value.
According to the company, Gemini 3 Flash can outperform Gemini 2.5 Pro while running three times faster. For reasoning-heavy tasks, it also uses around 30 percent fewer tokens, which can reduce total costs despite the higher per-token price.
Google describes Flash as a “workhorse” model designed for bulk and high-frequency tasks, making it especially attractive for companies operating at scale.
Intensifying Competition With OpenAI
Since the release of Gemini 3, Google says its API now processes over one trillion tokens per day, highlighting the intensity of the ongoing AI race.
Earlier this month, reports suggested that OpenAI issued an internal alert after ChatGPT traffic declined as Google’s consumer AI usage grew. In response, OpenAI released GPT-5.2 and a new image generation model, while also highlighting rapid growth in enterprise adoption and message volume.
Google has avoided directly framing the launch as a competitive move but acknowledged that the pace of innovation across the industry is accelerating.
According to the Gemini product leadership team, new models and evolving benchmarks are pushing all companies to improve faster and explore new ways to measure intelligence, efficiency, and real-world usefulness.