Google releases Gemini 3.7 Flash with enhanced agentic coding and lower API pricing
Google has introduced Gemini 3.7 Flash, a new general purpose AI model that improves coding, automation, document processing, and agentic performance while cutting API pricing compared with its predecessor. Google describes it as an updated “workhorse” model built around core optimizations and developer feedback.
The biggest gains are in coding and automation. Gemini 3.7 Flash scores 43.6% on FrontierCode 1.1 Main, up from 34.4%, and 65.3% on DeepSWE v1.1, compared with 49%. It also improves from 22% to 34% on the GDP.pdf document benchmark and from 17% to 30.4% on AutomationBench. The release comes only three weeks after Gemini 3.6 Flash, while the previously announced Gemini 3.5 Pro remains unreleased.
Gemini 3.7 Flash is available through the Gemini API, Google AI Studio, Antigravity, Android Studio, and Gemini Enterprise. Introductory pricing is $0.75 per million input tokens and $3.75 per million output tokens through the end of 2026, half the launch price of 3.6 Flash. In the Gemini app, it currently powers the Spark agent for AI Pro and Ultra subscribers, while the regular chatbot continues to use Gemini 3.6 Flash.
- Large Language Model (LLM) Tool
- Freemium
- Proprietary
Related news
External links
- Gemini 3.7 Flash: our most intelligent workhorse modelGoogle Blog • Official source
- Gemini 3.7 Flash launches three weeks after last model9to5Google
- Google announces Gemini 3.7 Flash just three weeks after previous releaseArs Technica
- Google joins the AI model price war with the new Gemini 3.7 FlashNeowin
Comments
3.7 is very good, ive been using it in antigravity and its noticeably stronger than 3.6
Clearly aimed at competing with OpenAI Luna, but is 3x more expensive (since it allows audio and images input), when Luna was already competing with all open-weight models, but is 3-10x more expensive (even with 50% off).
Most companies are replacing 90% of AI tasks with open-weight models, and even optimizing theses models and host them by themselves, making impossible to get profit from theses small proprietary models. But they're still pouring billions into it to reassure investors they're still AI companies.
Most companies are replacing 90% of AI tasks with open-weight models, and even optimizing theses models and host them by themselves, making impossible to get profit from theses small proprietary models. But they're still pouring billions into it to reassure investors they're still AI companies.
Typical AI Bubble Thing, they will still pour billions so that this circular economy of AI keeps of moving, after all it's showcase for funds and investors to raise more capital