GEMINI 3 FLASH
1M native context. Multimodal (text + image + audio + video). Native googleSearch grounding. $0.50/$3 per million tokens — the fast/cheap counterpart to Gemini 3 Pro.
Gemini 3 Flash is Google's fast-tier model in the Gemini 3 generation. It inherits the same 1M+ native context window and full multimodal stack as Gemini 3 Pro — text, image, audio, video — but at Flash latency and $0.50/$3 pricing instead of $2.50/$10. Quality is materially closer to Pro than the 2.x-era Flash was: it's the first Flash that's actually plausible for production frontier work, not just batch summarization. The case for: high-throughput multimodal pipelines, real-time chat over long context, vision QA at scale, anything that needs fresh information via googleSearch grounding. The case against: hairy coding (use Sonnet 4.6 or Opus 4.7), long-form reasoning that needs deeper chains (use Gemini 3 Pro or GPT-5.5 with high reasoning). In councils, Gemini 3 Flash slots in as the multimodal speed lane.
Three Gemini speed tiers, three jobs. Flash-Lite 3.1 ($0.25/$1.50) is the cheapest, best for free-tier traffic and bulk processing where any frontier-quality output is fine. Flash 3 ($0.50/$3) is the sweet spot — materially smarter than Flash-Lite, fast enough for chat, multimodal capable. Pro 3 ($2.50/$10) is the heavyweight — pick it when you need deep reasoning, not just speed.
Big jump in reasoning quality and instruction-following, much better tool use, and improved long-context coherence past 200K tokens. Price moved from $0.30/$2.50 (2.5 Flash) to $0.50/$3 (3 Flash) reflecting the capability bump.
Yes — natively. Upload a video file and ask about it. Same multimodal pipeline as Gemini 3 Pro, at Flash latency and cost.
$0.50 per million input tokens and $3.00 per million output tokens. Council AI bundles it inside monthly plan budgets — no separate Google billing.
Yes — coherence holds well into the high six figures. Beyond ~800K tokens you'll see some drift; Gemini 3 Pro holds together a bit further.
Pick Flash 3 for chat, vision, and any task where output quality matters. Pick Flash-Lite 3.1 for bulk processing where 2x the cost difference matters more than the quality gap.