Google introduced Gemini 3.7 Flash, positioning the model as a faster-moving option for coding, agents and knowledge work. The release arrived three weeks after Gemini 3.6 Flash and is available at temporary API prices that are scheduled to double at the start of 2027.
During the introductory period, Google set prices at $0.75 per million input tokens and $3.75 per million output tokens. From January 1, 2027, the stated rates are $1.50 and $7.50 respectively. The company described the launch price as half the original per-token price of Gemini 3.6 Flash.
Google reported higher results for the new model across several evaluations. On FrontierCode 1.1 Main, it listed a score of 43.6%, compared with 34.4% for version 3.6. DeepSWE v1.1 results rose from 49.0% to 65.3%. In Arena.ai’s WebDev Arena, Google reported Elo scores of 1,588 for 3.7 Flash and 1,538 for its predecessor.
The company also highlighted document and workflow tests. Gemini 3.7 Flash scored 34.0% on GDP.pdf, against 22.0% for 3.6 Flash, and 30.4% on AutomationBench, compared with 17.0%. These figures come from Google’s product announcement; the supplied evidence does not include independent replication or detailed test conditions.
Beyond benchmark scores, Google says the model follows instructions more closely, handles obstacles more effectively and makes fewer attempts before completing engineering tasks. It is also intended to generate more complete web applications from fewer prompts and to follow visual references such as screenshots or design systems. Those are vendor claims that users will need to assess against their own workloads.
Gemini Spark, Google’s agent product for AI Pro and Ultra subscribers in more than 160 countries, began using 3.7 Flash on the event date. Google said the change is intended to improve tasks involving Workspace tools, including consolidating files, preparing email drafts and updating project-status documents.
The release includes revised safeguards aimed at chemical, biological, radiological and nuclear risks as well as malicious cyber activity. Google directed developers to the model card for further information about those protections.
Gemini 3.7 Flash reflects a rapid update cycle in which price, tool use and software performance are central points of competition. Developers considering adoption must weigh the introductory price deadline, benchmark relevance and the practical reliability of the model in production. Google’s published comparisons indicate substantial improvement over 3.6 Flash, but real-world results will depend on prompts, integrations and evaluation methods.
The announcement does not disclose latency figures, context limits or availability by every API region. Those operational details, along with independent testing, will matter to teams comparing the release with other models. The short interval since version 3.6 also means adopters may need processes for evaluating frequent model changes before promoting them into established applications.
