Google DeepMind has today announced significant expansions to its Gemini 2.5 family, unveiling the stable releases of Geminiย 2.5 Pro and Geminiย 2.5 Flash alongside a preview of the allโnew Geminiย 2.5 FlashโLite model. These updates reflect Googleโs continued commitment to offering a spectrum of AI models that balance cost, speed, and performance for diverse workloads .
Stable Releases: Geminiย 2.5 Pro & Flash
On Juneย 17, 2025, Google marked the general availability of Geminiย 2.5 Pro and Geminiย 2.5 Flash. The Pro variant delivers maximum reasoning power and is tailored for highโcomplexity tasks such as advanced code generation, scientific analysis, and largeโscale data synthesis. In contrast, Geminiย 2.5 Flash offers a midโtier option optimized for everyday uses that demand low latencyโideal for chatbots, summarization, and content creation at scale.
Overview: Three Models in the Geminiโฏ-2.5 Family
| Model | Status | Strengths | Ideal Use Cases |
|---|---|---|---|
| Geminiโฏ2.5 FlashโLite (preview) | Preview | Fastest & cheapest; multimodal; controllable reasoning; tool-enabled | High-volume tasks like chatbots, summarization, search |
| Geminiโฏ2.5 Flash | Stable | Balanced: low latency, good reasoning, multimodal | Real-time conversations, customer support |
| Geminiโฏ2.5 Pro | Stable | Most capable: deep reasoning, huge context, multimodal | Research, complex coding, scientific tasks |
Geminiโฏ2.5 FlashโLite: Preview Highlights
Ultraโlow latency & cost savings:Designed for high-volume, real-time applications like translation, classification, and summarization. Boasts faster inference and lower cost per call compared to both 2.0 FlashโLite and the full Flash version .
Improved foundational performance: Outperforms earlier FlashโLite models across benchmarks in code generation, logic, math, multimodal reasoning, and science.
Cost and efficiency: FlashโLite pricing (preview): ~$0.10 per 1M input tokens and ~$0.40 per 1M output tokensโsignificantly cheaper than Flash ($0.30/$2.50) and Pro ($1.25/$10) .
Full Geminiโฏ-2.5 capabilities:
- Controllable Thinking: Users can set โthinking budgetsโ (token limits) to trade speed for depthโFlashโLite can toggle this on as needed.
- Multimodal Input: Supports text, image, audio, and video (including hour-long clips), with abilities to parse charts, UI, scenes, event summaries .
- Tool Integration: Includes Google Search, code execution and a million-token context window, matching the capabilities of Flash and Pro.
Positioning on the PriceโPerformance Curve
Google positions FlashโLiteโs high speed and low cost at the Pareto frontier, meaning itโs among the most costโefficient-yet-capable models worldwide (). In comparative evaluations, FlashโLite represents the best value: smart yet affordable .
About Flash and Pro
- Geminiโฏ2.5 Flash: Stable, low-latency, multi-modal thinking model. Positioned below Pro but roughly on par with GPT-4o in capability, with superior speed and cost efficiency ().
- Geminiโฏ2.5 Pro: Googleโs most advanced model. Renowned for handling hours-long video/audio, complex code and math, and huge-context reasoning. Also introduces selective โthinking budgetsโ and improved code quality to serve as a long-term stable flagship AI .
Deployment & Pricing
- Availability: All three models are accessible through Google AI Studio, Google Cloud Vertex AI, and the Gemini app .
- Cost structure (Vertex AI pricing from June 16, 2025):
- Pro: $1.25/1M input, $10/1M output (higher beyond 200K tokens)
- Flash: $0.15/1M input, $3.50/1M output in โthinkingโ modeโand includes 1,500 free grounded prompts daily ()
- FlashโLite (preview): ~$0.10/$0.40 per 1M tokens
Getting Started
CometAPI provides a unified REST interface that aggregates hundreds of AI modelsโunder a consistent endpoint, with built-in API-key management, usage quotas, and billing dashboards. Instead of juggling multiple vendor URLs and credentials.
Developers can accessย Gemini 2.5 Flash-Lite (preview) API throughย CometAPI, the latest models listed are as of the articleโs publication date. To begin, explore the modelโs capabilities in theย Playgroundย and consult theย API guideย for detailed instructions. Before accessing, please make sure you have logged in to CometAPI and obtained the API key.ย CometAPIย offer a price far lower than the official price to help you integrate.