Gemini 3.7 Flash arrives at half the price of 3.6 Flash
On 13 August, Google released Gemini 3.7 Flash, three weeks after Gemini 3.6 Flash. Flash is Google's line of cheaper, faster models, the ones developers put inside apps that make many calls a day. Google calls 3.7 Flash its most intelligent workhorse model yet for coding and agents.
The price is the headline. Until the end of the year, 3.7 Flash costs $0.75 per million input tokens and $3.75 per million output tokens, which Google describes as half the original price of 3.6 Flash.
What Google says got better
Google's numbers compare 3.7 Flash with 3.6 Flash. On DeepSWE v1.1, a test of fixing real software issues, it scores 65.3% against 49.0%. On FrontierCode 1.1 Main it goes from 34.4% to 43.6%. On AutomationBench, which checks whether a model can finish ordinary business workflows, it moves from 17.0% to 30.4%. On GDP.pdf, a test of reading complex documents, it moves from 22.0% to 34.0%.
Google also says the model asks for clarification when a request is unclear, follows instructions more closely and spends more effort planning steps and tool calls. In practice that should mean fewer retries when it runs as an agent, though those are Google's claims and the post does not show results from outside testers.
Where you can use it
Developers can call it through the Gemini API in Google AI Studio, in Android Studio and in Google Antigravity. Companies get it through Gemini Enterprise. For individuals, it now runs Gemini Spark, the personal agent for Google AI Pro and Ultra subscribers, which Google says can consolidate files, draft emails and update status documents. According to Google's support page, Spark is not offered in the European Economic Area, Switzerland, the United Kingdom or Nigeria.
What to do if you already use Flash
If an app of yours calls 3.6 Flash, switch the model name in a test copy and run the same prompts you run every day. Compare the answers and the bill for a week before you switch production. Write down that the price is introductory: Google says it holds through the end of the year, and the post does not say what it becomes after that.
What we think
Half the price with better coding scores is a good trade on paper. The comparison only goes against Google's own previous model, so the useful test is still your own prompts. We would also plan the budget as if the price goes up in January, because an introductory price is by definition temporary.
Sources