The new “workhorse” model outperforms competitors on benchmarks, enters Gemini app and search by default, and targets developers and enterprises with rapid, low-cost capabilities.
Gemini 3 Flash: A Major AI Step Forward
Google has officially released Gemini 3 Flash, a faster, more affordable AI model designed to handle high-volume, multimodal tasks—and it’s now the default model in the Gemini app and AI-powered search experiences globally.
- Replacing Gemini 2.5 Flash, the new model brings major improvements in performance, reasoning, and speed.
- It aims to serve as Google’s high-efficiency alternative in the escalating AI model wars with OpenAI.
“We really position Flash as more of your workhorse model… It actually allows for, for many companies, bulk tasks,” said Tulsee Doshi, Senior Director of Gemini Models at Google.
Benchmark Bragging Rights
Gemini 3 Flash posts impressive benchmark results, outperforming or matching top-tier models, including OpenAI’s GPT-5.2, in several key areas.
- Humanity’s Last Exam (no tools):
- Gemini 3 Flash: 33.7%
- GPT-5.2: 34.5%
- Gemini 3 Pro: 37.5%
- Gemini 2.5 Flash: 11%
- MMMU-Pro (multimodality + reasoning):
- Gemini 3 Flash: 81.2% (highest of all models tested)
These scores reflect the model’s strength in understanding text, images, audio, and video, as well as making intelligent decisions across domains.
Built for Multimodal Use Cases
Gemini 3 Flash is optimized for tasks that involve rich media inputs and visual reasoning:
- Upload a video (e.g., pickleball highlights) and get feedback or tips
- Submit a drawing and ask the model to identify it
- Analyze audio recordings, generate quizzes, or create interactive content
- Generate tables, images, or app prototypes using simple prompts
Google says this model is ideal for video analysis, data extraction, visual Q&A, and real-time workflows due to its speed and efficiency.
Developer and Enterprise Rollout
Gemini 3 Flash is already in use by enterprise partners such as:
- JetBrains
- Figma
- Cursor
- Harvey
- Latitude
It’s available via:
- Vertex AI and Gemini Enterprise platforms
- API preview for developers
- Antigravity, Google’s new AI-powered coding tool
Pricing: Fast and (Relatively) Cheap
While Gemini 3 Flash is slightly more expensive than its 2.5 predecessor, it’s significantly more efficient, especially for thinking-intensive tasks.
| Model | Input Tokens | Output Tokens |
|---|---|---|
| Gemini 2.5 Flash | $0.30 / 1M | $2.50 / 1M |
| Gemini 3 Flash | $0.50 / 1M | $3.00 / 1M |
- Uses 30% fewer tokens for many tasks compared to Gemini 2.5 Pro
- Three times faster than previous models
- Optimized for repeatable, large-scale workloads
Consumer Access and Feature Updates
- Gemini app users globally now default to Gemini 3 Flash
- Users can manually switch to Pro for advanced math and coding needs
- Gemini 3 Pro also rolling out more broadly in U.S. search
- Nano Banana Pro, an image model, is expanding to more U.S. users
Google vs. OpenAI: The Model Race Heats Up
The timing of this release appears strategic. Google’s AI product momentum is reportedly having a real impact:
- A “Code Red” memo was sent by Sam Altman at OpenAI after ChatGPT traffic declined, allegedly due to Gemini’s rising popularity.
- OpenAI quickly followed with GPT-5.2 and a new image model.
- Google, meanwhile, says it is processing over 1 trillion tokens daily via its APIs.
“All of these models are continuing to be awesome, challenge each other, push the frontier,” said Doshi. “It’s encouraging us to raise the bar.”
