Google Gemini 3 Flash undercuts Pro 4× – 3× faster multimodal runs

Google flipped the switch on Gemini 3 Flash today, and it’s quietly a big deal for anyone paying their own API bill. The new "Fast" brain across Gemini surfaces runs roughly 3× faster than Gemini 2.5 Pro and comes in at about a quarter of Gemini 3 Pro’s price: $0.50 per 1M input tokens, $3 per 1M output, plus $1 for audio. Benchmarks put Gemini 3 Flash (Reasoning) wedged between Gemini 3 Pro Preview and Claude Opus 4.5, so the headline is simple: near‑Pro quality at a 4× discount.

Google’s infra story backs that up. Context caching can shave up to 90% off repeated context, and the Batch API promises around 50% savings on async jobs like bulk script analysis or image review. Flash is already wired into the Gemini app (Fast mode), Search’s AI Mode, AI Studio, Vertex AI, Gemini Enterprise, the Gemini CLI, Android Studio, and Antigravity IDE, plus it’s now selectable in Perplexity alongside GPT‑5.2 and Opus 4.5. Creators are calling it their mobile daily driver because the latency finally matches constant checking, outlining, and sketching.

The pattern is clear: Pro‑class models are becoming the "weekend deep work" tools, while cheaper, fast variants like Gemini 3 Flash take over the every‑hour storyboarding, copy, and UX iteration loops