Google Updates Gemini 1.5 Pro, Halves Its API Pricing
Google has rolled out an improved version of Gemini 1.5 Pro with better math and code performance, and cut API pricing by up to half to attract more developers.
Google updated Gemini 1.5 Pro this week, its large-context-window model, adding performance improvements and a steep price cut to the API developers use to integrate the model into their own applications.
The company said the new version of Gemini 1.5 Pro performs better at math, generating longer code, and following complex instructions compared to the previously available version. Gemini 1.5 Flash, the lighter, cheaper variant built for fast, high-volume tasks, also got quality tweaks.
The price cut is the headline news
Beyond the quality improvements, the move the industry will actually notice is the pricing. Google has lowered the cost of using Gemini 1.5 Pro through the API for prompts up to 128,000 tokens: the price per million input tokens drops from $7 to $3.50, and output pricing goes from $21 to $10.50. In other words, cut in half. For contexts beyond that 128,000-token threshold — where Gemini 1.5 Pro touts its window of up to one million tokens, one of the largest on the market — prices remain higher, though they've also been adjusted downward.
A token is roughly the smallest unit of text a language model processes: a short word or part of a longer one. The more tokens a model can handle in its context, the more text it can "read" at once — an entire book, hours of transcribed video, entire codebases — without losing track of what came before.
Why Google is making this move now
Since OpenAI launched GPT-4o in May with lower prices than its predecessor, and Anthropic did the same with its Claude 3.5 family, the race to offer intelligence at a lower cost per token has become one of the most visible battlegrounds among major AI labs. Price per token is now one of the most direct selling points for companies choosing between models when building products: the cheaper the API call, the more feasible it becomes to deploy a model at scale — in assistants, internal search tools, or customer service systems that handle thousands of queries a day.
Google has spent months arguing that Gemini 1.5's long context — its ability to handle massive amounts of information at once — is its key differentiator against rivals like GPT-4o or Claude 3.5 Sonnet, whose context windows are notably smaller. Making that capability cheaper to access is meant to turn that technical edge into real adoption among developers, who until now may have seen the cost of exploiting very long contexts as a barrier to entry.
What changes for current Gemini users
For developers already integrating Gemini 1.5 Pro into their applications, the change is automatic: there's no need to update code or switch model versions to benefit from the new pricing. For those using Gemini through Google AI Studio or Vertex AI, Google's cloud platform for businesses, the quality upgrade also arrives with no extra steps required.
The move comes at a time when Gemini is competing for space against ChatGPT and Claude both among developers and the general public, and when Google itself has been expanding access to its Gemma family of open models, positioned as a lightweight alternative for those who want to run models locally or customize them.
The open question is how long this price war can continue without eating into the margins of the companies waging it. For now, the main winners are the people building products with these models: paying less per token processed, when operating at a scale of millions of daily calls, adds up to a cost difference that's anything but symbolic.
