2 min read AI-generated

Gemini 3.7 Flash: Google Aims at Coding Agents – and Halves the Price

Copy article as Markdown

Three weeks after 3.6 Flash, Google sends its next model for coding and agents into the ring. It's said to beat comparable models from Anthropic and OpenAI on nine benchmarks – at half the price.

Featured image for "Gemini 3.7 Flash: Google Aims at Coding Agents – and Halves the Price"

Google unveiled Gemini 3.7 Flash yesterday – a model tuned specifically for coding and agentic workflows. The pace alone is worth noting: only three weeks separate 3.6 Flash and 3.7 Flash. Google is picking up its update tempo, and you can feel it.

What the model can do

The pitch is clear: a low-cost option for companies building autonomous systems. Agents that plan tasks, operate tools, and push through multi-step workflows with little human input. Exactly the field where Claude Code and Codex play too.

On benchmarks, Google puts numbers on the table. On FrontierCode 1.1 Main, 3.7 Flash hits 43.6%, up from 34.4% for its predecessor. On DeepSWE v1.1 it’s 65.3% versus 49.0%. Google says the model beats comparable models from Anthropic and OpenAI on nine benchmarks. Take in-house comparisons with a grain of salt – but the jump over its own predecessor is clear.

The real lever: the price

More interesting than any benchmark is the price sheet. Through the end of the year, 3.7 Flash costs $0.75 per million input tokens and $3.75 per million output tokens – half of 3.6 Flash. Starting January 1, 2027, it rises to $1.50 and $7.50 respectively. So Google isn’t giving anything away permanently, but it’s leaning hard on price in the short term to win users.

The model is broadly available: in GitHub Copilot, Google Antigravity, AI Studio, Android Studio, and the Gemini Enterprise Agent Platform. If you work in Copilot, you can use it right inside the IDE.

My take

The cadence is the real news. Three weeks between two models – that’s not a model cycle anymore, that’s a software release. And the direction is unmistakable: fast, cheap, good enough for agents. Google is attacking exactly what makes Claude Code strong.

For us as users, this competition is a gift. When Google halves the price to hold its own against Anthropic and OpenAI, whoever builds is the one who benefits in the end. I’m staying with Claude because the day-to-day experience works for me – but I’m still going to look at 3.7 Flash. A model that delivers solid coding numbers at half the price isn’t something you just ignore. And the pressure it puts on every provider makes the coming months interesting.


Sources: