Published: 14 August 2026
Google released Gemini 3.7 Flash on 13 August 2026, a fast, low cost model that the company positions as its most capable workhorse model for coding and agents. Gemini 3.7 Flash arrives 23 days after Gemini 3.6 Flash, keeps the same 1,048,576 token context window, and launches at an introductory price of 0.75 dollars per million input tokens and 3.75 dollars per million output tokens. On Google’s own evaluations the model scores 65.3 percent on DeepSWE v1.1, against 49.0 percent for Gemini 3.6 Flash.
What can Gemini 3.7 Flash do?
Gemini 3.7 Flash is Google’s fast, low cost tier, tuned this cycle for software engineering, agent workflows and multi step execution rather than raw reasoning. Google describes it as its most intelligent workhorse model yet for coding and agents, and every headline benchmark it published is a software engineering or automation test. The model accepts text, images, video, audio and PDF as input, and Google says it can generate functional layouts and feature complete applications in fewer prompts while following a screenshot, image or design system closely.
The model exposes three thinking levels, low, medium and high, and defaults to medium. The minimal setting that worked on some earlier Gemini models now returns an error. Because output tokens include thinking tokens, that setting is the main lever on cost per call. Google also says the model can clarify intent, adjust when it hits a roadblock, and put more effort into multi step planning and tool calls, which is the behaviour it is targeting for agent use.
Gemini 3.7 Flash benchmarks and technical specs
Google published five benchmark pairs against Gemini 3.6 Flash, all from its own evaluations.
- Gemini 3.7 Flash scores 43.6 percent on FrontierCode 1.1 Main against 34.4 percent for Gemini 3.6 Flash,
- 65.3 percent on DeepSWE v1.1 against 49.0 percent,
- Elo of 1588 on WebDev Arena against 1538,
- 34.0 percent on GDP.pdf against 22.0 percent,
- 30.4 percent on AutomationBench against 17.0 percent. The AutomationBench figure is worth reading at its absolute level as well as its delta: at 30.4 percent the model still fails roughly seven in ten multi step automation tasks.
The model card carries four scores that did not appear in the announcement.
- Gemini 3.7 Flash reaches 97.0 percent on GDM-MRCR v2 at 128k,
- 85.8 percent on Terminal-bench 2.1,
- 47.9 percent on OSWorld-2.0 for computer use.
- On the harder Terminal-bench 3.0 the score drops to 14.9 percent, which is a useful reminder of how much a benchmark number depends on which edition is quoted.
On specifications, Gemini 3.7 Flash has a 1,048,576 token input context window and a 65,536 token output limit, both unchanged from Gemini 3.6 Flash, and a knowledge cutoff of March 2026. The third party site Artificial Analysis scores the model 56 on its Intelligence Index against 52 for Gemini 3.6 Flash, and ranks it first of 186 models on output speed at 340.1 tokens per second.
Gemini 3.7 Flash pricing and how it compares to GPT-5.6 Luna and DeepSeek V4-Flash
Gemini 3.7 Flash costs 0.75 dollars per million input tokens and 3.75 dollars per million output tokens. Google’s own footnote states that this is introductory pricing that expires on 31 December 2026, after which the rate rises to 1.50 dollars per million input tokens and 7.50 dollars per million output tokens, exactly what Gemini 3.6 Flash cost at its launch. Context caching is priced at 0.075 dollars per million tokens plus 0.50 dollars per million tokens per hour of storage. Google has also moved Gemini 3.6 Flash onto the same introductory rate, so the two models cost the same today and the upgrade decision turns on capability rather than price.
Against rival models Gemini 3.7 Flash is not the cheapest option in its tier. OpenAI’s GPT-5.6 Luna is listed at 0.20 dollars per million input tokens and 1.20 dollars per million output tokens, and DeepSeek V4-Flash at 0.14 dollars and 0.28 dollars. Gemini 3.7 Flash does undercut Claude Haiku 4.5, listed at 1.00 dollars and 5.00 dollars, on both sides. Artificial Analysis puts the blended price of Gemini 3.7 Flash at 0.58 dollars per million tokens, half the 1.16 dollars it recorded for Gemini 3.6 Flash.
Where Gemini 3.7 Flash is available
Developer access started on 13 August 2026. Gemini 3.7 Flash is live through the Gemini API in Google AI Studio, in Android Studio, in Google Antigravity for agent first workflows, and across the Gemini Enterprise Agent Platform and app. The model identifier is gemini-3.7-flash.
Consumer access is narrower. In the Gemini app the model reaches users through Spark, Google’s always on personal agent, which requires a Google AI Pro or Ultra subscription. Google says Spark is available in more than 160 countries, and a footnote to the announcement excludes the European Economic Area, the United Kingdom, Switzerland and Nigeria. Readers in those regions cannot reach Gemini 3.7 Flash through the Gemini app today, whatever they pay for AI Pro or Ultra, though API access is unaffected. Google gave no timeline for those markets. The release ships with updated Frontier Safety safeguards covering chemical, biological, radiological and nuclear misuse and cyber offence.
For more information you can visit the official Google Gemini 3.7 Flash announcement.
