Google on Wednesday released Gemini 3.7 Flash, its latest AI model optimized for coding and agentic tasks, just three weeks after launching Gemini 3.6 Flash and at half the introductory cost. The release, announced in a blog post by Senior Director Tulsee Doshi, marks the fastest update cycle in the Flash model series and signals Google’s aggressive push to dominate the AI agent market.
The model delivers significant gains in software engineering benchmarks. On FrontierCode 1.1 Main, 3.7 Flash scored 43.6 percent compared to 34.4 percent for 3.6 Flash. On DeepSWE v1.1, it reached 65.3 percent versus 49.0 percent. The model also outperformed its predecessor on Arena.ai’s WebDev Arena leaderboard with an Elo score of 1,588 versus 1,538, and on the GDP.pdf benchmark for complex document comprehension at 34.0 percent versus 22.0 percent.
Cheaper, Faster, More Capable
Google priced 3.7 Flash at $0.75 per million input tokens and $3.75 per million output tokens through the end of 2026, roughly half the launch price of 3.6 Flash. Starting January 1, 2027, pricing will rise to $1.50 and $7.50 per million tokens respectively. The model supports a context window of up to 1 million tokens and maximum output of 64,000 tokens.
The release comes amid a broader acceleration in Google’s AI release cadence. Gemini 3.6 Flash launched in late July, 3.5 Flash in late June, and 3.5 Pro is still in testing. Google is releasing new models at a pace that outstrips rivals OpenAI and Anthropic, both of which typically space major releases months apart.
Gemini Spark Agent Gets Upgrade
The new model also powers Gemini Spark, Google’s 24/7 personal AI agent launched at I/O in May for Pro and Ultra subscribers across more than 160 countries. Google says Spark with 3.7 Flash can now handle complex multi-skill workflows more efficiently, including consolidating files, drafting emails, and updating status documents in Google Workspace.
Early adopters including Box, Databricks, LangChain, and Pydantic reported significantly improved performance over 3.6 Flash. Google’s blog post featured testimonials from 13 companies praising the model’s ability to complete production-ready coding tasks with fewer retries and less manual oversight.
The release also includes updated safety safeguards against misuse in chemical, biological, radiological, and nuclear domains, as well as cyber offense. Google published a detailed model card with benchmark results and known limitations, noting the model’s knowledge cutoff of March 2026.
Developers can access 3.7 Flash through Google AI Studio, the Gemini API, Android Studio, and Google Antigravity. Enterprise customers can deploy it via the Gemini Enterprise Agent Platform and Gemini Enterprise app on Google Cloud.
Sources: Google Blog; Ars Technica; The Verge; DeepMind model card
discussion