Gemini 3.8 Flash

Gemini 3.8 Flash (2 September 2026) is Google's most capable Flash model, aimed at long-horizon software engineering and autonomous agents — and, at $0.75 per million input tokens, roughly a thirteenth of what the current flagships charge. That rate is introductory and expires on 31 December 2026.

Best for

• Bulk production traffic where a flagship is overkill
• Agentic work with a large, stable cached prefix
• Long-horizon coding and multi-step tool use
• Anything where cost per task decides the architecture

How it compares — and the India angle

• The price doubles on 1 January 2027 — $0.75/$3.75 becomes $1.50/$7.50, cached input $0.075 becomes $0.15, and cache storage $0.50 becomes $1.00 per million tokens an hour. Google publishes the date itself; budget at the January rate and treat the difference as a four-month discount.
• Dropping to Gemini 3.7 or 3.6 Flash does not dodge the rise: all three sit on one rate card and double on the same day.
• Output pricing includes thinking tokens, so a model set to think longer bills more at an identical advertised rate. Artificial Analysis measured $0.58 per index task here against $0.40 on 3.7 Flash — about 30% more output tokens for the same sticker price.
• Indian teams: at $0.75 in and $3.75 out this is the cheapest sensible default for volume, and the free tier is genuinely usable — but free-tier content is used to improve Google's products, which rules it out for client work.