Google’s Gemini 2.5 Flash: powerful AI at a bargain price

Google just released Gemini 2.5 Flash, and I’m honestly impressed by how much performance they’ve packed into such an affordable AI model. Let me break down why this matters for businesses like yours.

What makes Gemini 2.5 Flash special?

This new model stands out for three key reasons:

• It’s surprisingly affordable
• It handles multiple types of content (text, audio, images, and video)
• It can process up to a million tokens of context (that’s massive)

The model introduces a unique “thinking mode” that lets you control how deeply it reasons through complex problems. You can literally set a “thinking budget” to determine how many tokens (up to 25,000) the AI should use when tackling your requests.

The pricing is where it gets interesting

Let’s talk numbers:

• 15¢ per million tokens of input
• 60¢ per million tokens of output
• $3.50 per million reasoning tokens

To put this in perspective, competitor models like Anthropic’s Claude Sonnet charge $3 per million output tokens, and others charge up to $15 per million. Gemini 2.5 Flash delivers comparable performance at a fraction of the cost.

How does it perform?

Surprisingly well. In benchmark tests, Gemini 2.5 Flash:

• Scores just below OpenAI’s newest GPT-4o Mini
• Outperforms Claude 3.7 with thinking enabled
• Shows dramatic improvement over previous Gemini versions (jumping from 27.5% to 78% on