Gemini 3 Flash
2025-12-17 · Publisher : Google · Domain : Artificial Intelligence
Google launches Gemini 3 Flash, a model that combines the Pro-level reasoning of Gemini 3 with the speed and cost of a Flash: USD 0.50 input and USD 3 output per million tokens.
What changes
- Outperforms Gemini 2.5 Pro on several benchmarks while being three times faster (Artificial Analysis measurement) and cheaper.
- Modulates its thinking: uses 30% fewer tokens than 2.5 Pro on average on typical traffic.
- Becomes the default model of the Gemini app (replacing 2.5 Flash) and of AI Mode in search.
- Strong performance in agentic coding and multimodal reasoning.
### Specifications - Identifier: `gemini-3-flash-preview`; context of 1 million input tokens and 65,536 output tokens (API documentation). - Inputs: text, image, video, audio, PDF. - Availability: Gemini API (AI Studio), Antigravity, Gemini CLI, Android Studio, Vertex AI, Gemini Enterprise, Gemini app. ### Pricing | Model | Input (USD / 1M tokens) | Output (USD / 1M tokens) | |---|---|---| | Gemini 3 Flash | 0.50 (text, image, video); 1.00 (audio) | 3.00 | Batch and Flex at 0.25 / 1.50 according to the current pricing page. ### Benchmarks | Benchmark | Score | |---|---| | GPQA Diamond | 90.4% | | Humanity's Last Exam (without tools) | 33.7% | | MMMU-Pro | 81.2% | | SWE-bench Verified | 78% | Scores published by the vendor. ### Key takeaways The most attractive price-performance ratio in the Gemini 3 range: a default option for agents and high-traffic assistants.
Links
Tags : LLM, Google, Gemini, Flash, Gemini 3