Enterprise

Google ships Gemini 3.7 Flash three weeks after 3.6, halves API price, and still won't date Gemini 3.5 Pro

Alphabet's workhorse model gets a coding-and-agents refresh at $0.75 per million input tokens through year-end, even as the flagship Pro release keeps slipping.

By Asher Glick · Enterprise software desk · 2026-08-14 · 3 min read

Photo: Unsplash / Chris Ried — Lines of source code displayed on a developer's monitor

Google released Gemini 3.7 Flash on Thursday, exactly three weeks after 3.6 Flash, and cut the API price to half the previous rate: $0.75 per million input tokens and $3.75 per million output through December 31, 2026. The flagship Gemini 3.5 Pro, which Google told partners in July was “coming soon,” is still nowhere on the calendar.

Read those two facts together and the strategy becomes legible. Alphabet is shipping the cheap, high-frequency workhorse on an accelerated cadence while the marquee Pro release keeps slipping, a sequencing choice that lets Google occupy the developer news cycle without answering the question investors actually want answered.

The benchmarks are real. 3.7 Flash posts 43.6% on FrontierCode 1.1 Main against 3.6’s 34.4%, and 65.3% on DeepSWE v1.1 against 49.0%. Google claims wins over comparable models from Anthropic and OpenAI on nine benchmarks, including a 6% margin over Claude Sonnet 5 and 9.3% over GPT-5.6 Terra on the GDP.pdf business-document evaluation. Per the model card, it’s the same transformer-based mixture-of-experts architecture derived from Gemini 3 Pro, tuned for the agentic loop: Google says the model “better adapts to roadblocks, clarifies intent when needed, and follows instructions with greater fidelity.”

Distribution is what you’d expect. 3.7 Flash routes through the Gemini Enterprise Agent Platform, the Gemini Enterprise app, and Spark, Google’s 24/7 personal agent for AI Pro and AI Ultra subscribers.

The context Google isn’t foregrounding is last week’s leadership overhaul at Google DeepMind. Demis Hassabis stepped aside in favor of his deputy Koray Kavukcuoglu, and the two original technical co-leads of Gemini quit to co-found a startup. Bloomberg notes investors have been watching for 3.5 Pro as a read on whether DeepMind can keep pace with Anthropic and OpenAI. Axios reports the company is already training Gemini 4 and encouraged by early results, raising the possibility that 3.5 Pro gets skipped entirely for a Gemini 4 Pro release.

Half-price Flash is a real product decision. It’s also a narrative-management exercise: ship the tier you can ship, price it aggressively, and let the benchmarks do the talking while the org chart resets behind the curtain.

Sources