Google has launched Gemini 3 Flash, a fast AI model that provides cutting-edge intelligence at much lower costs than earlier versions. This release increases access to AI features across Google products, including developer tools and daily apps. It performs better than Gemini 2.5 Pro on most tests, being three times faster and up to 70% cheaper.
Key Features
Gemini 3 Flash combines professional-grade reasoning with quick speed, excelling in multitasking like video analysis, image captioning, and coding assistance. It scores high on benchmarks, including GPQA Diamond (90.4%), Humanity’s Last Exam (33.7%), and MMMU Pro (81.2%), competing with larger models. Developers appreciate its low latency for real-time tasks, such as in-game assistants and A/B testing.
Performance Benchmarks
Gemini 3 Flash surpasses Gemini 2.5 Pro in 18 out of 20 tests, with three times faster response times (218 tokens per second) and 30% fewer tokens used on average tasks. On SWE-bench Verified, it scores 78%, outperforming even Gemini 3 Pro for coding tasks.
| Benchmark | Gemini 3 Flash | Gemini 2.5 Pro | Improvement |
|---|---|---|---|
| GPQA Diamond | 90.4% | Lower | Frontier-level |
| MMMU Pro | 81.2% | Lower | State-of-the-art |
| SWE-bench Verified | 78% | Lower | Outperforms Pro |
| Output Speed | 218 tokens/sec | ~70-80 tokens/sec | 3x faster |
Pricing and Availability
Gemini 3 Flash costs $0.50 per million input tokens and $3 per million output tokens, offering significant savings over earlier models. It's being rolled out worldwide as the default in the Gemini app, AI Mode in Search, Vertex AI, and in developer tools like Google AI Studio and Antigravity.
Developer and User Impact
For developers, it allows quick updates in high-frequency workflows, with companies like JetBrains and Figma already using it. Everyday users benefit from quick multimodal analysis, such as turning videos into action plans or creating apps using voice commands. This positions Gemini 3 Flash as a major innovation for scalable AI applications.
0 Comments