Gemini 3.7 Flash Reaches General Availability

A New Workhorse for Coding and Agents

Google introduced Gemini 3.7 Flash on August 13, 2026, calling it "our most intelligent workhorse model yet for coding and agents." The release lands roughly three weeks after Gemini 3.6 Flash, continuing Google's rapid iteration cadence on the Flash line. According to Google's model card, 3.7 Flash is a refinement of 3.6 Flash built on algorithmic improvements to the model's core reasoning rather than a fresh pretraining run — a faster, lower-cost way to ship capability gains.

The model is available immediately through the Gemini API under the identifier gemini-3.7-flash, and supports a 1,048,576-token input context window with up to 65,536 output tokens. It accepts text, image, video, audio, and PDF input, and supports capabilities including code execution, function calling, structured outputs, computer use (in preview), search grounding, Google Maps grounding, URL context, caching, and three configurable thinking modes (low, medium, high).

Benchmark Gains in Coding and Agentic Work

Google reports substantial improvements across the benchmarks it uses to track coding and agent performance. On FrontierCode 1.1 Main, a production-code-quality eval, Gemini 3.7 Flash scores 43.6% compared to 34.4% for Gemini 3.6 Flash. On DeepSWE v1.1, a long-horizon software-engineering benchmark, it reaches 65.3%, up from 49.0%. The model also posts the top score in Google's comparison table on WebDev Arena, an Elo-based web-development benchmark, moving from 1538 to 1588 — reflecting more functional layouts and complete applications generated with fewer prompts.

Knowledge-work benchmarks show similar gains: GDP.pdf, an expert-level PDF comprehension eval, rises from 22.0% to 34.0%, and AutomationBench, a private enterprise-workflow benchmark, climbs from 17.0% to 30.4%. Google attributes these improvements to better multi-step planning, stronger instruction-following fidelity, and a model that adapts to obstacles and asks for clarification rather than stalling or looping — all of which reduce the manual oversight and retries developers previously had to build around.

Introductory Pricing Through 2026

Alongside the capability jump, Google is cutting the entry price for the model roughly in half. Through December 31, 2026, Gemini 3.7 Flash is priced at $0.75 per million input tokens and $3.75 per million output tokens. Standard pricing of $1.50/$7.50 per million tokens takes effect on January 1, 2027. Google is also offering the model through Batch API, Flex inference, and Priority inference consumption tiers for workloads with different latency and cost tradeoffs.

Availability

Gemini 3.7 Flash is rolling out across Google's developer and consumer surfaces simultaneously: Google AI Studio, Android Studio, and Google Antigravity for developers; the Gemini Enterprise Agent Platform for business customers; and Spark, the agentic feature in the Gemini app, for Google AI Pro and Ultra subscribers in more than 160 countries. GitHub also added the model to GitHub Copilot for Pro, Pro+, Max, Business, and Enterprise subscribers on the same day. Google says early enterprise partners — including Box, Databricks, Harvey, Hebbia, LangChain, and Pydantic — have already validated the model's performance and cost profile ahead of the public rollout.

Safety Updates

Google notes that Gemini 3.7 Flash ships with updated safety guardrails specifically targeting chemical, biological, radiological, and nuclear (CBRN) misuse and cyber-offense scenarios, alongside the model's broader capability improvements.