Artificial Intelligence

Gemini 3.7 Flash Halves Its Price Until January

Google launched Gemini 3.7 Flash on August 13 at half its listed price through the end of 2026, when the rate doubles. Coding scores jumped in just 23 days, but the discount has a calendar, not a condition.

Artificial IntelligenceBy Servola Tech Desk2026-08-144 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Gemini 3.7 Flash Halves Its Price Until January

Key takeaways

Google released Gemini 3.7 Flash on August 13, 2026, an introductory tier priced at 0.75 dollars per million input tokens and 3.75 dollars per million output tokens - exactly half of the 1.50 and 7.50 dollar rate that takes effect on January 1, 2027, according to Google's own announcement of the model.
The model's coding scores moved sharply in the 23 days since the prior Flash release: FrontierCode 1.1 Main rose from 34.4 percent to 43.6 percent and DeepSWE v1.1 from 49.0 percent to 65.3 percent, according to Google's published benchmark table.
Gemini 3.7 Flash is Google's fifth Flash-tier model since May 2026 (3.5 Flash, 3.5 Flash-Lite, 3.5 Flash Cyber, 3.6 Flash, now 3.7 Flash), a release cadence of roughly one new tier every three to four weeks that both 9to5Google and VentureBeat separately flagged as unusually fast for a frontier lab.
Any business that wires 3.7 Flash's launch price into an agent workflow today is budgeting against a rate that doubles in under five months, with Google giving no public signal on whether a 3.8 Flash tier, and a fresh discount, arrives before or after that January 1 reset.

Three Flash Tiers in Ten Weeks

Google shipped Gemini 3.7 Flash on August 13, 2026, just 23 days after Gemini 3.6 Flash, according to Google's own DeepMind blog post announcing the model. It is the fifth Flash-tier release since May 2026, following 3.5 Flash, 3.5 Flash-Lite, 3.5 Flash Cyber (a security-hardened variant), and 3.6 Flash - a cadence both 9to5Google and VentureBeat separately described as unusually fast, even for a company that ships as often as Google does.

The model is live now in the Gemini app under the Spark feature for Pro and Ultra subscribers, in Google Antigravity, in AI Studio, in Android Studio, and in the Gemini Enterprise Agent Platform and app, per Google's launch materials. That is a broader simultaneous rollout than most single-tier launches get, suggesting Google wants developer and enterprise traffic on the new model immediately rather than staged over weeks.

A Discount With a Fixed End Date

The headline number is 0.75 dollars per million input tokens and 3.75 dollars per million output tokens, an introductory rate Google says holds through December 31, 2026. From January 1, 2027, the published rate doubles outright to 1.50 dollars per million input tokens and 7.50 dollars per million output tokens - not a taper, a hard reset on a fixed calendar date that is already public.

Google is also pitching the model on capability, not just price: its own materials claim Gemini 3.7 Flash beats Anthropic's Claude on certain business-workflow benchmarks, according to reporting from Tech Times, and the release adds hardened safeguards against chemical, biological, radiological, nuclear, and cyber misuse. A model that is cheaper, faster to adopt, and framed as safer is a strong pitch during the exact window when developers are choosing which vendor's API to standardize on - which is also the window before the price doubles.

What Actually Improved

Beyond the two headline coding benchmarks, Google's own table shows WebDev Arena Elo moving from 1538 to 1588, GDP.pdf (a knowledge-intensive document benchmark) rising from 22.0 percent to 34.0 percent, and AutomationBench climbing from 17.0 percent to 30.4 percent. Those are agentic-workflow and document-reasoning tests, not just raw coding puzzles, which is consistent with Google's stated focus on coding and agent tasks for this release.

Google describes the model as better at adapting to roadblocks mid-task, clarifying intent when instructions are ambiguous, and executing multi-step plans with less manual correction - the kind of incremental reliability gain that matters more to a team running an agent in production than a single benchmark headline does, because it is the difference between a workflow that needs a human to babysit it and one that mostly does not.

The Renewal Date Nobody Is Budgeting For

The practical risk here is not the model, it is the calendar. Any European or UK team that standardizes an agent workflow on Gemini 3.7 Flash's launch pricing this autumn is implicitly building a budget that assumes today's rate, when Google has already published the date it stops being true. January 1, 2027 is not a rumor or an analyst estimate - it is the vendor's own stated reset date, which makes it one of the easiest cost risks in AI procurement to actually plan for, and one of the easiest to ignore because it is five months out.

This is not a Google-specific pattern either: the same week Gemini 3.7 Flash launched, DeepSeek separately announced a peak-hour pricing mechanism taking effect August 16 that raises its own flagship rates several times over. Launch-week pricing across frontier AI vendors is increasingly a promotional number with an expiry date attached, not a stable unit cost - so the sensible move for any procurement team is to calendar the reset date the day the contract starts, not the week before the bill changes.

Frequently asked questions

Will Gemini 3.7 Flash's price stay at 0.75 dollars per million tokens?

No. Google set that rate as an introductory price running through December 31, 2026. From January 1, 2027, the published rate doubles to 1.50 dollars per million input tokens and 7.50 dollars per million output tokens, according to Google's own announcement.

How does Gemini 3.7 Flash compare to OpenAI's or Anthropic's coding models?

Google's own materials claim it beats Claude on certain business-workflow benchmarks, and independent outlets note its FrontierCode and DeepSWE scores now sit close to larger frontier models, though neither OpenAI nor Anthropic had published a head-to-head response as of this writing.

Where can businesses access Gemini 3.7 Flash?

It is live in the Gemini app under the Spark feature for Pro and Ultra subscribers, in Google Antigravity, AI Studio, Android Studio, and the Gemini Enterprise Agent Platform, according to Google's launch announcement.

The interesting number in this launch is not 0.75 dollars, it is January 1, 2027. Any team that lets a procurement decision run on autopilot past that date will find out the hard way that a launch price and a real price are two different contracts.

Google DeepMindGemini 3.7 FlashAI pricingcoding modelsagentic AIGoogle

Open-Source AI Adds a Full Day to Cyclone Warnings

Google DeepMind open-sourced WeatherNext, a peer-reviewed AI model that gives 3-day cyclone forecasts the accuracy 2-day forecasts had before - a full extra day of warning, free to download on GitHub.

4 min read

Artificial Intelligence

DeepMind's CEO Steps Aside as Three Scientists Exit

Demis Hassabis moves to chairman and Alphabet chief scientist as Jeff Dean, Sanjay Ghemawat and Oriol Vinyals leave to found Discovery Loop. Alphabet stock fell nearly 4 percent, and Koray Kavukcuoglu now runs Gemini day to day.

5 min read

Decisions

The Third Pivot Was the Warning Magic Leap Missed

Magic Leap filed a WARN notice for 193 layoffs effective 1 October 2026, its fourth strategic pivot after $4 billion raised. The lesson is the pattern, not the pivot.