Google officially launched Gemini 3.5 Flash on May 19, ending the tech industry’s long-running compute subsidy phase. For the last couple of years, major tech companies subsidized the hardware costs of artificial intelligence, giving developers access to incredibly cheap, capable models. That era is over. The new Gemini 3.5 Flash costs $1.50 per million input tokens and $9.00 per million output tokens.
Those numbers are an exact 300% price increase over the model it replaces. The previous iteration, Gemini 3 Flash, ran at just $0.50 for input and $3.00 for output. Developers and startups built countless applications around those throwaway prices, assuming the cost of raw machine intelligence would simply continue to drop. Now, major AI labs are testing how much businesses are actually willing to pay for these tools.
The price jump breaks the traditional routing logic that developers use to balance speed and cost. The $1.50 and $9.00 price tags push the budget Flash tier uncomfortably close to the cost of Google’s premium Gemini 3.1 Pro, which sits at $2.00 for input and $12.00 for output. This narrow price gap forces enterprise teams to rethink how they distribute high-volume coding and text classification workloads across their systems.
Not every company is following Google’s strategy. While the search giant raises costs, open-weight competitors like DeepSeek are breaking the trend of falling budget AI prices by aggressively cutting their own rates to win over frustrated customers.
How the Gemini 3.5 Flash Pricing Squeezes Enterprise Teams
When the cost of processing a million tokens triples overnight, the financial math for high-volume AI applications breaks down completely. Startups rely on budget models like Flash to process enormous datasets, summarize thousands of documents, and handle basic customer service chats. A 3x cost increase on those repetitive tasks directly eats into profit margins.
The narrow gap between Gemini 3.5 Flash and Gemini 3.1 Pro also changes the value proposition of Google’s AI tier system. If developers only save fifty cents per million tokens by downgrading to Flash, many will simply default to the smarter Pro model for reliability, or abandon the Google ecosystem entirely for cheaper open-source alternatives. DeepSeek’s aggressive price cuts show that the budget AI market is fracturing between premium proprietary ecosystems and race-to-the-bottom open weights.