After unveiling Gemini 2.5 Flash, the smaller model in the 2.5 series, at the Google Cloud Next ’25 event roughly a week ago, Google has now made it available through Google AI Studio and Vertex AI, disclosing its pricing and capabilities.
Google has priced the Gemini 2.5 Flash slightly higher than its predecessor, the Gemini 2.0 Flash, albeit with added Thinking capabilities in the former but not in the latter. The input cost stands at $0.15 per 1M token, output cost at $0.30 per 1M token, and $0.60 if reasoning features are used. This is in contrast to the pricing of 2.0 Flash, which had input and output costs of $0.10 and $0.40 respectively.
The benchmark scores of the 2.5 Flash have significantly improved compared to the 2.0 Flash, closely aligning with its competitor model, the DeepSeek R1, but at a much lower cost (DeepSeek R1 input cost $0.55, output cost $2.19).
Google defines Gemini 2.5 Flash as “Our most cost-efficient thinking model,” highlighting its price-performance ratio in the market (refer to the accompanying graph). Currently, Google offers 4 models to choose from based on expertise level (2.5 Pro being the most advanced, while 2.0 Flash Lite is the most economical), but in terms of performance-cost ratio, Gemini models emerge as the most worthwhile among competitors.
Another standout feature of running 2.5 Flash is the thinking budget system that sets a predefined number of tokens used in thinking mode in advance (the more tokens used, the higher the quality of responses, but also at a higher cost), thereby managing costs right from the start.
TLDR: Google introduces Gemini 2.5 Flash with enhanced capabilities, competitive pricing, and a unique thinking budget system for cost-effective and efficient performance.
Leave a Comment