Google launches Gemini 3.5 Flash Low to fix developer quota issues

Google has launched a new Gemini 3.5 Flash (Low) model on its Antigravity development platform to address user complaints about simple tasks using too many tokens. The company also reset all free and paid Gemini plan quotas to give users enough capacity for the week.

Varun Mohan, Google DeepMind director in charge of Antigravity, said the team heard users were unhappy with token usage on simple tasks. The new version cuts costs by adjusting the model's reasoning effort level, not by cutting corners on system prompts or context compression.

Official data shows the Low version uses about 45% fewer tokens than the Medium version, and often performs better than the older Gemini 3 Flash on software engineering tasks. This move helps ease recent frustration over tighter quotas.