.png){kind=small}
On December 17, Google announced Gemini 3 Flash, delivering frontier-level AI performance at 3x the speed of its predecessor.
At $0.50 per million input tokens, it achieves 78% on SWE-bench Verified and is comparable to Gemini 3 Pro on key multimodal tasks (81.2% MMMU Pro).
The model uses 30% fewer tokens than previous versions while scoring 90.4% on PhD-level reasoning, is now free for consumers via the Gemini app, and is starting to roll out as the default model powering AI Mode in Google Search, with developer access through Gemini API, Vertex AI, and Google Antigravity.
Companies like Figma, JetBrains, and Bridgewater are already integrating it for production applications.
Real-World Impact
Since the Gemini 3 launch, Google has processed over 1 trillion tokens daily, validating the model's production readiness at scale. The combination of speed and multimodal capabilities enables real-time video analysis, data extraction, and visual processing, supporting workflows that previously required larger models.
For developers, this means agentic coding workflows and interactive applications that were once resource-intensive are now more accessible to small teams.
The economic shift is decisive: sophisticated AI features that were previously expensive are now available at $0.50 per million tokens, fundamentally changing who can build advanced AI applications.
.png)
.svg)
.png)
.png)
