Gemini 3.7 Flash Shines in Tests Despite Some Flaws

Google's Gemini 3.7 Flash model has undergone significant upgrades compared to its predecessor, Gemini 3.6 Flash. One of the most notable improvements is its ability to perform tasks that were previously beyond its capabilities.

In a test, Gemini 3.7 Flash was given a prompt to build a playable browser game in just two minutes and 13 seconds, which it successfully accomplished. This task stumped Gemini 3.6 Flash three weeks earlier, highlighting the significant progress made by the new model.

Gemini 3.7 Flash's performance was also put through various other tests, including a creative writing test where it was tasked with generating a story that obeyed specific structural rules. While the resulting story showed promise, it still suffered from some issues typical of AI-generated content, such as overuse of adjectives and clichés.

In contrast to more expensive models like GPT-5.6 Sol, Gemini 3.7 Flash's performance was considered competent but not exceptional. Its price, however, is significantly lower at $0.75 per million input tokens until December 31, before increasing to $1.50 on January 1.