Google launches Gemini 3 Flash as default model in Gemini app

Model Overview

Google introduced Gemini 3 Flash, a new AI model that builds on the Gemini 3 system released the previous month. Designed to be both faster and more cost‑effective than its predecessor, Flash is positioned as the workhorse model for everyday tasks. The company announced that Gemini 3 Flash will replace Gemini 2.5 Flash as the default model in the Gemini app worldwide, while users can still select the Pro variant for specialized math and coding queries.

Performance and Capabilities

On benchmark evaluations, Gemini 3 Flash outperformed Gemini 2.5 Flash by a large margin and matched or exceeded other leading models in several measures. For example, on the Humanity’s Last Exam benchmark, Flash achieved a score of 33.7%, compared with 11% for the older Flash version and 34.5% for a competing model. In multimodal reasoning tests, the model posted an 81.2% score, surpassing rivals.

Flash supports a range of multimodal inputs. Users can upload short videos to receive coaching tips, submit sketches for identification, or provide audio recordings for analysis and quiz generation. The model’s improved intent understanding enables richer visual answers that incorporate images and tables.

Pricing and Availability

Google disclosed pricing of $0.50 per million input tokens and $3.00 per million output tokens for Gemini 3 Flash, slightly higher than the prior version but justified by the model’s speed and token efficiency. The model uses about 30% fewer tokens on average for thinking tasks, potentially lowering overall costs for bulk operations.

Developers can access Gemini 3 Flash through Vertex AI, Gemini Enterprise, and a preview API. The model is also integrated into Google’s Antigravity coding tool and the Gemini app’s prototype creation features.

Enterprise Adoption

Several technology firms, including JetBrains, Figma, Cursor, Harvey, and Latitude, have begun using Gemini 3 Flash in their products and services. Google highlighted that the model’s speed and affordability make it well‑suited for video analysis, data extraction, and visual question‑answering workflows that require rapid, repeatable execution.

Future Outlook

Google’s rollout of Gemini 3 Flash reflects its broader strategy to compete aggressively in the generative AI space. By positioning Flash as the default model for consumer‑facing applications and offering the Pro version for high‑performance needs, Google aims to capture a larger share of both everyday and enterprise AI workloads. Executives emphasized that continuous benchmarking, new evaluation methods, and rapid model iteration will keep the ecosystem dynamic and competitive.