On August 13, 2026, Google launched Gemini 3.7 Flash, its new AI model built for coding and agents, with entry pricing cut in half compared to the previous version. Only three weeks separate this release from Gemini 3.6 Flash, launched in late July. For an SME leader tracking AI API bills or running automations, this pace and this price cut are worth understanding precisely before deciding whether to adjust a budget or a toolset.
In brief
- Google launched Gemini 3.7 Flash on August 13, 2026, described as its "most intelligent workhorse model" for coding and agents (source: Google DeepMind).
- Entry pricing is cut in half: $0.75 per million input tokens and $3.75 per million output tokens, against a standard rate that doubles on January 1, 2027 ($1.50 and $7.50).
- On the DeepSWE software-development benchmark, the score rises from 49.0% to 65.3%; on FrontierCode, from 34.4% to 43.6% (source: Google DeepMind).
- On AutomationBench, which measures the ability to carry out typical office tasks (invoicing, email sorting, reporting), the score rises from 17.0% to 30.4%.
- The model is available immediately in Google AI Studio, Gemini Enterprise, and the Gemini Spark app, in more than 160 countries.
- For an SME, the takeaway is not to chase every new model release, but to make use of a limited pricing window before the scheduled increase in January 2027.
What Google announced on August 13, 2026
Gemini 3.7 Flash is the third release in the "Flash" line in just over a month, following Gemini 3.5 Flash-Lite and Gemini 3.6 Flash, launched on July 21, 2026. It is not Google's most powerful model (Gemini Pro remains above it), but it is built to run at volume: agents chaining dozens of calls per task, continuously generated code, automated workflows.
According to Google DeepMind, the model's developer, the gains come both from developer feedback and from algorithmic changes targeted at three use cases: software engineering, web development, and knowledge work (finance, law, life sciences). The model is already deployed for developers (Google AI Studio, Android Studio, Google Antigravity), for businesses (Gemini Enterprise Agent Platform), and for individuals subscribed to Gemini Spark.
The measured gains on coding and agents
The scores published by Google DeepMind show a clear improvement on the tasks most useful to a business: writing and fixing code, navigating a site to complete a task, or automating an office workflow.
| Metric | Gemini 3.6 Flash | Gemini 3.7 Flash |
|---|---|---|
| DeepSWE v1.1 | 49.0% | 65.3% |
| FrontierCode 1.1 Main | 34.4% | 43.6% |
| AutomationBench | 17.0% | 30.4% |
| WebDev Arena (Elo) | 1,538 | 1,588 |
| Price (per M tokens, input / output) | Not disclosed / $7.50 | $0.75 / $3.75 (launch price, through 12/31/2026) |
Sources: Google DeepMind (official blog post, August 13, 2026), MarkTechPost, VentureBeat.
These remain lab benchmark scores, not a guarantee of results for any specific use case. But the jump on AutomationBench, which simulates concrete administrative tasks, is the most telling signal for an SME: it's exactly the kind of task an internal automation (invoice sorting, template replies, data extraction) is meant to handle.
A price cut, but for how long?
The $0.75 / $3.75 per million token rate is not permanent. It is a launch price, valid through December 31, 2026. Starting January 1, 2027, it doubles: $1.50 for input and $7.50 for output, the same level as Gemini 3.6 Flash's output price.
Key takeaway
An automation built today on Gemini 3.7 Flash will cost twice as much per call starting in January 2027, barring a new price cut announced by Google before then. Before building a critical process on this rate, it's worth checking that the math still works out at the standard price, not just the promotional one.
This pricing pattern is not new among major AI providers: attractive introductory rates, tightened once adoption is established. For an SME, the right approach remains the same one already recommended for previous Gemini launches: use the low price to test and calibrate a use case, without pinning a multi-year budget to it.
An unusual release pace
Mid-2026
Gemini 3.5 Flash-Lite
July 21, 2026
Gemini 3.6 Flash
August 13, 2026
Gemini 3.7 Flash
Three versions of the "Flash" model in just over a month: Google is visibly focusing its efforts on its cheapest, most agent-friendly line, while the release pace of its flagship Gemini Pro model has slowed, according to several trade publications (InfoWorld). In other words, the current race is less about the "most intelligent model in absolute terms" and more about the model most cost-effective to run continuously, a criterion that speaks directly to an SME.
What this concretely changes for an SME
Recalculate the cost of an existing automation
Test before rolling out broadly
Plan for the end of the promotional rate
In practice, an SME automating email sorting, quote-draft generation, or code review doesn't need to switch tools overnight. But checking whether a cloud provider or AI vendor already offers Gemini 3.7 Flash as an option, and comparing the real cost over a month of testing, is a simple habit worth adopting this fall.
FAQ
What does Gemini 3.7 Flash concretely change for my AI bill?
If your automations already use a Gemini Flash model, switching to version 3.7 can reduce the cost per task through December 31, 2026, thanks to the launch rate of $0.75 / $3.75 per million tokens. The gain depends on the token volume your current workflows consume.
Will Gemini 3.7 Flash's price stay this low?
No, barring a new announcement from Google. The current rate is a launch price valid through December 31, 2026. It then doubles ($1.50 input, $7.50 output), matching Gemini 3.6 Flash's output price.
Does Gemini 3.7 Flash replace Gemini 3.6 Flash?
Yes, Google presents Gemini 3.7 Flash as the latest version of its "workhorse" line for coding and agents. Gemini 3.6 Flash remains available, but new deployments are encouraged to use version 3.7.
Do I need to switch AI providers to benefit from this price?
Not necessarily. If your tech stack already runs on the Google ecosystem (Google AI Studio, Vertex AI, Gemini Enterprise), switching to Gemini 3.7 Flash is typically a simple configuration change, not a full tool migration.
In conclusion
Gemini 3.7 Flash's price cut confirms a trend observed over recent months: the cost of agentic AI keeps falling, making automations profitable that weren't a year ago. But a launch price is not a guaranteed price, and an SME should test before scaling, then budget for the increase announced for 2027. To follow the previous Google price cut, read our article on Gemini 3.6 Flash and the cost of AI, or explore our Mag resources to stay current on news useful to SME leaders.
