Gemini 3.7 Flash Pricing: The "Half Price" Launch Has an Expiration Date

Google shipped Gemini 3.7 Flash on August 13, 2026, just 23 days after Gemini 3.6 Flash, and every headline said the same thing: half the price of its predecessor. That's technically true and also not the whole story. The discount is temporary, a rival model already matched the "old" price, and a chunk of Europe can't reach the model at all right now.

Here's what the launch-day coverage left out.

Gemini 3.7 Flash pricing has an expiration date

Gemini 3.7 Flash costs $0.75 per million input tokens and $3.75 per million output tokens. Google's own footnote spells out what happens next: that rate is introductory and expires on December 31, 2026. Starting January 1, 2027, it jumps to $1.50 input and $7.50 output, per million tokens, exactly what Gemini 3.6 Flash cost when it launched in July.

There's a second wrinkle that makes the "50% cheaper" headline weaker than it sounds. Google quietly moved Gemini 3.6 Flash onto the same $0.75/$3.75 introductory rate. As of today, the two models cost identically the same. The upgrade case has to rest on capability, not price, because there isn't a price difference to point to anymore.

Gemini 3.7 Flash beats Claude Haiku 4.5 on price, which is a real win. But [GPT-5.6 Luna](https://questloops.com/blog/chatgpt-s-free-tier-just-got-a-massive-upgrade-gpt-5-6-luna-explained) and DeepSeek V4-Flash both undercut it by a wide margin, and both of those stay flat in January while Google's rate doubles. If you're picking a model purely on cost per token, Google's Flash line isn't the cheapest chair at the table, introductory pricing or not.

Where the coding gains are real

The benchmark jump is the actual news here, more than the price. On FrontierCode 1.1, Gemini 3.7 Flash scores 43.6%, up from 34.4% for 3.6 Flash. On DeepSWE v1.1, a long-horizon software engineering test, it climbs from 49.0% to 65.3%, a 16-point gain in three weeks. AutomationBench, which measures multi-step workflow automation, nearly doubles from 17.0% to 30.4%.

Independent measurement backs up the direction, if not the drama. Artificial Analysis puts Gemini 3.7 Flash at 56 on its Intelligence Index versus 52 for 3.6 Flash, and ranks it first out of 186 models on raw output speed at 340.1 tokens per second.

One number worth sitting with: that AutomationBench score of 30.4% still means the model fails roughly seven out of ten multi-step automation tasks. The improvement is large. The absolute level is still low. Google led with the delta, not the level, which is standard practice but worth noticing.

The part almost nobody reported: who can't use it

Google says Spark, its always-on personal agent, is available to subscribers "in over 160 countries." What most launch-day coverage skipped is the next line: the rollout excludes the European Economic Area, the United Kingdom, Switzerland, and Nigeria.

If you're in the EU, UK, or Switzerland, Gemini 3.7 Flash isn't reachable through the Gemini app right now, no matter what tier you're paying for. The API is unaffected, so developers in those regions can still build on the model directly. It's only the consumer Spark surface that's locked out, and Google hasn't given a timeline for fixing that.

Should you switch?

If you're already on Gemini 3.6 Flash, yes, switch now. The two models cost the same today and 3.7 Flash wins every published benchmark, some by a wide margin. There's no reason to stay on the older one.

If you're picking a budget model from scratch for agentic coding work, Gemini 3.7 Flash earns a real evaluation, particularly given that DeepSWE result. If you're picking purely on price, GPT-5.6 Luna or DeepSeek V4-Flash cost meaningfully less, and neither one has a price hike scheduled for January. For a broader look at where the current model lineup stands on cost, see our breakdown of [how to cut your AI agent's API bill](https://questloops.com/blog/how-to-cut-your-ai-agent-s-api-bill-token-compression-smart-routing-and-when-a-gateway-pays-for-itself).

FAQ

**Is Gemini 3.7 Flash free?** No. In the Gemini app it only runs through Spark, which needs a Google AI Pro or Ultra subscription. Developers can reach it through the paid Gemini API, though a free API tier does list the model with lower rate limits.

**When does the introductory price end?** December 31, 2026. The rate rises from $0.75/$3.75 to $1.50/$7.50 per million tokens on January 1, 2027.

**Why can't I use it in Europe?** Spark, the only consumer surface running the model, currently excludes the EEA, UK, Switzerland, and Nigeria. The API route is unaffected.

Google is now shipping a new Flash model roughly every three weeks, and each one resets the pricing conversation. Whatever you decide today, put a reminder on your calendar for late December, since the price change is real and it's easy to miss.