Google releases Gemini 3.1 Flash-Lite at $0.25 per million tokens

$0.25 per million tokens. Let that sink in. We ran a customer support bot last quarter that cost maybe $12/month on GPT-4o Mini — that same workload on Flash-Lite is now closer to $1.50. Google isn't competing anymore, they're trying to make everyone else's business model not work.

Here's the thing though — pricing this aggressive isn't charity. Google wants you locked in before you notice it. Lock-in wrapped in a discount is still lock-in. (We've been through this with AWS credits and we know exactly how it ends.)

Honestly, for commodity inference — summarization, classification, structured extraction — this is the obvious call right now. Flash-Lite is also 2.5x faster than previous Flash, so cost and latency objections disappear at the same time. That's a hard combo to argue against.

What I'd actually watch: how long this pricing holds. OpenAI's been in the same race. When o3-mini dropped, everyone recalculated their stacks. This is the same move, different jersey. Google has the infrastructure margins to sustain a price war longer than most — that's not nothing.

For us? We're swapping Flash-Lite into every low-stakes inference call this week. API surface is compatible enough that it's an afternoon, not a sprint.