Google shipped Gemini 3.7 Flash on Thursday, three weeks after Gemini 3.6 Flash, at $0.75 per million input tokens and $3.75 per million output tokens, half the introductory cost of its predecessor. The rate expires December 31, 2026. What the company didn’t ship, and wouldn’t date, is Gemini 3.5 Pro, the flagship it told partners in July would arrive “soon.”

The cadence tells its own story. Flash releases are compounding while the Pro line stalls, and the pricing cut reads less like generosity than a bid to keep developers on the platform while the top of the stack lags.

On benchmarks, 3.7 Flash posts 43.6 percent on FrontierCode 1.1 Main against 34.4 percent for 3.6 Flash, and 65.3 percent on DeepSWE v1.1 against 49.0 percent. Google says it beat comparable Anthropic and OpenAI models, including Claude Sonnet 5 and GPT-5.6 Terra, on nine benchmarks, and cites a 34 percent score on the GDP.pdf document-comprehension test. Context window holds at 1 million tokens, output caps at 64,000.

“It better adapts to roadblocks, clarifies intent when needed, and follows instructions with greater fidelity,” said Tulsee Doshi, senior director of product management at Google. Gemini Spark, the consumer agent, now runs on 3.7 Flash for Google AI Pro and Ultra subscribers across more than 160 countries.

The Pro silence is louder in context. Last week, per Reuters, Demis Hassabis stepped aside atop Google DeepMind for his deputy Koray Kavukcuoglu, and two original technical co-leads of Gemini left to co-found a startup. Axios reports Gemini 4 is already in training, with early results Google calls encouraging.

A workhorse ships on time, cheaper. The flagship doesn’t ship at all. The org chart moves underneath both.

Sources