IMPLICATOR
Google released Gemini 3.7 Flash on Thursday with API prices of $0.75 per million input tokens and $3.75 per million output tokens through year-end, half Gemini 3.6 Flash's original launch rates. Google says the model follows instructions more closely. The offer gives developers a defined testing window to learn whether higher coding and agent scores lower the cost of completed work before token rates double.
What Changed
- Google released Gemini 3.7 Flash three weeks after Gemini 3.6 Flash.
- Introductory API rates are $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026.
- Gemini 3.7 Flash reached 65.3% on DeepSWE v1.1, up from 49.0% for its predecessor.
- The rates double on January 1, 2027, while Google's own table still shows several workload wins for GPT-5.6 Terra.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Coding scores rise
The release arrived Aug. 13, three weeks after Gemini 3.6 Flash. On FrontierCode 1.1 Main, a measure aimed at production code quality, Gemini 3.7 Flash scored 43.6% against 34.4% for its predecessor in the launch results. DeepSWE v1.1, which tests longer software-engineering tasks, rose to 65.3% from 49.0%.
Google says the changes improve debugging and first-pass code accuracy.
The clearest external measurement is WebDev Arena, where blind user preference votes produced a 1,588 Elo score for Gemini 3.7 Flash, compared with 1,538 for Gemini 3.6 Flash.
Business-workflow tests moved too. GDP.pdf, an evaluation of complex-document handling, increased to 34.0% from 22.0% in Google's August release materials. AutomationBench, which measures completion of business processes, increased to 30.4% from 17.0%.
The comparison stays mixed
GPT-5.6 Terra leads it on DeepSWE, both reported Terminal-bench versions and OSWorld-2.0. On CharXiv Reasoning without tools, Gemini 3.7 Flash scored 84.5%, down from 85.2% for Gemini 3.6 Flash in the same comparison.
Google's comparison table shows GPT-5.6 Terra at 69.6% on DeepSWE, against 65.3% for Gemini 3.7 Flash. It shows Terra at 87.4% versus 85.8% on Terminal-bench 2.1, 20.8% versus 14.9% on Terminal-bench 3.0, and 50.2% versus 47.9% on OSWorld-2.0. Those workload-specific results do not amount to a single aggregate rank. The useful comparison for a team depends on the specific workload it intends to run inside its own repositories and workflows.
FrontierCode and DeepSWE are commercial benchmark suites and Google presented the results in its launch materials. WebDev Arena is based on blind user votes, which makes it an outside preference measure.
AI moves fast. Keep up.
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
The published benchmarks do not establish cost per successfully completed task inside a developer's own repository or workflow. A cheaper token can still produce an expensive result if an agent needs more retries, longer context or more human correction. Teams will have to measure those costs against their own logs.
Capacity and access
Gemini 3.7 Flash accepts text, images, audio and video across a one-million-token context window and can return up to 64,000 output tokens. It is available through the Gemini API, Google AI Studio, Android Studio, Antigravity, Gemini Enterprise and the paid Gemini Spark agent.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
That input and output capacity matches Gemini 3.6 Flash, keeping the release focused on the model's behavior rather than a larger context allowance.
The regular Gemini chatbot did not use Gemini 3.7 Flash at launch. Individual access initially runs through Spark for Google AI Pro and Ultra subscribers, while developers and companies can reach it through Google's hosted products. There are no open weights, so the release does not offer a self-hosted option.
The price changes in January
Through Dec. 31, Google charges $0.75 per million input tokens and $3.75 per million output tokens. Those introductory rates are half Gemini 3.6 Flash's original launch prices and apply during the evaluation window.
On Jan. 1, 2027, the rates become $1.50 per million input tokens and $7.50 per million output tokens. By then, developers will need an answer from their own repositories and workflow logs: What did each successfully completed task cost?
Frequently Asked Questions
What is Gemini 3.7 Flash?
Gemini 3.7 Flash is Google's hosted workhorse model for coding, agent workflows, web development and knowledge work. It accepts text, images, audio and video across a one-million-token context window.
How much does Gemini 3.7 Flash cost?
Through December 31, 2026, Google charges $0.75 per million input tokens and $3.75 per million output tokens. On January 1, 2027, those rates rise to $1.50 and $7.50.
How did Gemini 3.7 Flash perform on coding benchmarks?
Google reported 43.6% on FrontierCode 1.1 Main and 65.3% on DeepSWE v1.1, up from 34.4% and 49.0% for Gemini 3.6 Flash.
Does Gemini 3.7 Flash lead every benchmark?
No. Google's comparison table places GPT-5.6 Terra ahead on DeepSWE, Terminal-bench 2.1, Terminal-bench 3.0 and OSWorld-2.0.
Where is Gemini 3.7 Flash available?
The model is available through the Gemini API, Google AI Studio, Android Studio, Antigravity, Gemini Enterprise and the paid Gemini Spark agent. Google did not release open weights.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
