Gemini 3.7 Flash Pricing Explained

Gemini 3.7 Flash costs $0.75 per 1M input tokens and $3.75 per 1M output tokens, an introductory rate that Google will double on January 1, 2027. That single sentence covers what most people searching for Gemini 3.7 Flash pricing need, but the introductory window changes the math on any budget you build now. Google launched the model in August 2026 as its workhorse option for coding and agents. This guide breaks down the current rates, what caching saves you, when the price rises, and how the cost compares against the model's measured performance. If you are weighing whether to build on it, the numbers below are the ones that matter.

Gemini 3.7 Flash costs $0.75 input and $3.75 output per million tokens

Gemini 3.7 Flash is priced at $0.75 per 1M input tokens and $3.75 per 1M output tokens on the Gemini API paid tier. Google positions it as a workhorse for coding and agents, and the low rates are built for high-volume production use. One detail matters for reasoning-heavy work: the output rate includes thinking tokens, so the model bills at the output price for the tokens it generates while working through a problem.

These rates cover the standard text path. Audio, image, and video inputs can bill differently, so price those separately if your build leans on non-text input.

Gemini 3.7 Flash API pricing, paid tier. Pricing as on August 2026.

Google extended the same introductory rate to Gemini 3.6 Flash, so the two models sit at identical pricing during the introductory window. If you are choosing between them on cost alone, there is no gap to weigh right now.

The introductory price doubles on January 1, 2027

Gemini 3.7 Flash pricing is temporary at its current level, and the increase is not small. On January 1, 2027, the input rate rises from $0.75 to $1.50 per 1M tokens, and output rises from $3.75 to $7.50 per 1M tokens. That is a straight doubling of both meters.

The introductory discount gives teams several months to test the model in production before the standard rate applies. For a high-volume coding agent or a document-processing pipeline, the difference between the two rates can reshape a monthly bill, so it is worth modeling both numbers rather than budgeting on the introductory rate alone. A workload that costs $2,000 a month now would cost $4,000 at the standard rate on the same token volume.

Context caching cuts input costs by about 90 percent

Context caching drops cached input to $0.075 per 1M tokens during the introductory period, a 90% reduction against the $0.75 base input rate. Caching is useful when you send the same large context repeatedly, such as a system prompt, a codebase, or a reference document that stays constant across many calls. You pay the reduced rate on the cached portion rather than the full input rate each time. Like the base rates, cached input doubles on January 1, 2027, to $0.15 per 1M tokens.

For agentic workloads that replay a long instruction set on every step, caching is often the single largest lever on cost. The savings scale with how much of your input is stable across calls, so a workflow with a fixed 100K-token context and short per-call queries benefits far more than one that sends fresh input every time.

Gemini 3.7 Flash is inexpensive relative to its measured intelligence

Gemini 3.7 Flash delivers strong measured performance for its price tier, which is the part of the pricing story that raw token rates miss. On the independent Artificial Analysis Intelligence Index, Gemini 3.7 Flash scores 56 in its high-reasoning configuration, placing it well above the median for models in its class. Artificial Analysis runs models on the same harness, so the figure is comparable across vendors rather than self-reported.

Independent performance and cost, high-reasoning configuration. Source: Artificial Analysis, August 2026.

The cost-per-task figure is where the value shows. Artificial Analysis measures the weighted cost to run its full evaluation suite, and Gemini 3.7 Flash comes in at $0.40 per task, a fraction of what frontier models charge to reach a similar or moderately higher score. If you want to see how it stacks against a heavier sibling, our 3.7 Flash vs 3.1 Pro breakdown covers that gap in detail. Google's own reported coding and agent benchmark gains over 3.6 Flash are larger still, though those are vendor-reported figures and await independent confirmation. For a value-tier model, scoring 56 while billing well below the class median on cost per task is the reason the pricing has drawn attention.

Where you can access Gemini 3.7 Flash

Gemini 3.7 Flash is available through the Gemini API via Google AI Studio, with the same rates applying whether you call it directly or through Google's developer tooling. The model ID is gemini-3.7-flash. It is also offered on the Gemini Enterprise Agent Platform and the Gemini Enterprise app for larger deployments, and through Gemini Spark, Google's personal agent, for Google AI Pro and Ultra subscribers.

A free tier exists for prototyping through AI Studio, with rate limits and the standard condition that free-tier usage may be used to improve Google's products. Production traffic runs on the paid tier at the rates above. If you already build across models, the same token-based billing logic applies here as with the rest of the Gemini 3.x line, which is worth keeping in mind if you compare it against a predecessor like Gemini 3.6 Flash.

Beyond the model pricing

Gemini 3.7 Flash pricing rewards teams that plan around two dates: the introductory rate today and the doubled standard rate from January 2027. For most builds, the practical takeaway is to model both, lean on context caching where your input is stable, and treat the $0.40 cost-per-task figure as the real signal of value rather than the headline token rates alone. It is a genuinely inexpensive model for the intelligence it delivers, and the introductory window makes the next few months the cheapest time to test it.

If your goal is to ship an application rather than manage token math, Emergent lets you build production-grade full-stack apps by describing what you want, with access to Gemini, Claude, and GPT models through a single Universal LLM Key. You pay for what you use, skip the API wiring, and deploy on your own domain. Get started with Emergent and turn a described idea into a running product.