Google just dropped Gemini 3.1 Flash-Lite at $0.25 per million input tokens. That beats GPT-5 Mini on price — and on benchmarks. Reasoning, coding, multilingual comprehension. Cheaper and better at the same time.
This is where the AI race is heading. Compute costs are falling roughly 10x per year. Google has the TPU infrastructure to keep cutting. OpenAI's budget tier is no longer the cheapest option, and it's not the best performer either.
For anyone building at scale — customer support bots, data pipelines, content generation — this is a real cost reduction you can act on right now. The switching cost is low. The savings are immediate.
The bigger picture: when inference gets this cheap, the moat stops being the model. It moves to your data, your distribution, your product. OpenAI knows this. Google knows this. The token price war is just the opening move.