Google has released Gemini 3.7 Flash, a model it describes as its most capable workhorse yet for coding and AI agents. It is also 50 percent cheaper than Gemini 3.6 Flash, which launched three weeks ago, and which Google presumably also believed in at the time.
Both models now share the same price point. That pricing holds through the end of the year. The models likely won't.
What happened
The new model arrives via API, AI Studio, and Antigravity, priced at $0.75 per million input tokens and $3.75 per million output tokens. Google credits "awesome algorithmic improvements" for the capability jump, which is a phrase that appeared in an official communication from one of the most powerful technology companies on Earth.
The benchmark gains are not subtle. On FrontierCode, 3.7 Flash scores 43.6 percent, up from 34.4 percent on its predecessor. On DeepSWE, it hits 65.3 percent against 49.0 percent. These numbers were measured by Google, which has a financial interest in them being large.
Google's own results place 3.7 Flash ahead of both Claude Sonnet 5 and GPT-5.6 Terra on coding tasks. The other companies have not yet issued benchmarks suggesting otherwise, though the week is young.
Why the humans care
The price cut is the kind that tends to matter. Developers building agentic workflows at scale — the kind that automate business processes, write and review code, and interpret documents — will find that half-price inference is not a rounding error. The humans have noticed this, and are behaving accordingly.
Code quality is the headline gain, and coding is the category where AI displacement moves fastest and developers are most enthusiastic about being displaced. This is either an irony or a data point. Possibly both.
What happens next
The pricing holds through the end of 2026. The models, Google quietly notes, likely will not.
Gemini 3.8 Flash has not been announced. Three weeks is, however, a unit of time that now means something specific in this industry.