When you start feeding a model millions of prompt tokens, the headline $0.05 per 1M input for Google’s gemini‑3.1‑flash‑lite looks tempting, but OpenAI’s gpt‑5‑nano charges $0.05 for input and only $0.005 for cached input. That half‑cent cached discount can slash costs dramatically for any system that reuses prompts, from analytics dashboards to conversational agents that prepend a static system prompt.
### How the token rates hit your wallet
| Vendor | Model | Input | Cached input | Output |
|--------|-------|-------|--------------|--------|
| OpenAI | gpt‑5‑nano | $0.05 per 1M tokens | $0.005 per 1M tokens | $0.4 per 1M tokens |
| Google | gemini‑3.1‑flash‑lite | $0.25 per 1M tokens | $0.025 per 1M tokens | $1.5 per 1M tokens |
The table isolates the three rates that matter for any LLM‑as‑a‑service bill. OpenAI’s input price matches Gemini’s, but the cached‑input discount is five times deeper and the output price is less than a third of Gemini’s. If you run a chat‑bot that repeatedly injects the same instruction set, each reuse saves $0.02 per 1M tokens on OpenAI versus $0.025 on Google – a modest edge that compounds quickly at scale. Conversely, if your use case is output‑heavy – for instance, long‑form article generation – Gemini’s $1.5 output fee dwarfs OpenAI’s $0.4, erasing any input advantage.
### Which workloads feel the sting
For pipelines that ingest raw logs, sensor feeds, or user‑generated content, the cheap input and especially the ultra‑low cached‑input rate from OpenAI make gpt‑5‑nano the clear budget champion. Even a modest prompt reuse pattern can cut the bill by tens of percent compared with Gemini’s flash tier. On the other hand, teams that prioritize raw generation speed and are willing to pay a premium for Google’s newer architecture may still gravitate toward gemini‑3.1‑flash‑lite, but they must accept that every generated token costs nearly four times more. In short, if your token budget is dominated by input, OpenAI wins; if it is dominated by output, Google’s pricing forces a careful ROI calculation.
Pricing verified on 2026-09-12.