Google AI Releases Gemini 3.7 Flash Coding and Agent Model Starting at $0.75

Google has officially announced the launch of Gemini 3.7 Flash, a new artificial intelligence model tailored specifically for software engineering and agentic workflows. Access to the model is provided exclusively through hosted environments and enterprise channels, including the Gemini API, Google AI Studio, Google Antigravity, Android Studio, the Gemini Enterprise Agent Platform, and the Gemini Enterprise application. Everyday consumers can access it via the Gemini Spark tier on Google AI Pro and Ultra plans, as Google has chosen not to release any open-weight versions for this rollout.

Regarding code generation and software development capabilities, the model demonstrates notable improvements across multiple benchmark evaluations:

• FrontierCode 1.1 Main measuring production code quality scores 43.6%, outperforming the 34.4% achieved by 3.6 Flash
• DeepSWE v1.1 long-horizon software engineering evaluation reaches 65.3%
• WebDev Arena posts an Elo rating of 1588 compared to 1538 previously, securing the top position in Google's comparison table
• GDP.pdf expert PDF comprehension evaluation increases from 22.0% to 34.0%
• AutomationBench private enterprise workflow set records 30.4%, surpassing Claude Sonnet 5 at 10.7% and GPT-5.6 Terra at 23.6%
• Long-context retrieval on GDM-MRCR v2 at 128k achieves 97.0%

Stock photo for illustration only, not from the actual event

Despite these strengths, competing models still maintain leads in certain specific categories. GPT-5.6 Terra leads in DeepSWE at 69.6%, Terminal-bench 2.1 at 87.4%, Terminal-bench 3.0 at 20.8%, and OSWorld-2.0 at 50.2%. Furthermore, CharXiv Reasoning experiences a regression, scoring 84.5% without tools compared to 85.2% for 3.6 Flash. On the Artificial Analysis Intelligence Index, 3.7 Flash records a score of 56, trailing slightly behind both GPT-5.6 Terra and Muse Spark 1.2 at 57.

Pricing for Gemini 3.7 Flash starts at an introductory rate of $0.75 per 1 million input tokens and $3.75 per 1 million output tokens, which remains active through December 31, 2026. Starting January 1, 2027, the rates will increase to $1.50 and $7.50, respectively. In Google's benchmark cost breakdown, Claude Sonnet 5 is priced at $2.00/$10.00 and GPT-5.6 Terra at $2.00/$12.00. Based on an 80/20 input-output mix, this translates to a blended cost of $1.35 per 1 million tokens today, compared to $3.60 for Sonnet 5 and $4.00 for GPT-5.6 Terra.

Source: MarkTechPost