When you start paying for millions of tokens, the gap between Google’s flash‑focused Gemini 3.5 Flash‑lite and OpenAI’s GPT‑4o mini becomes the decisive factor in your cloud‑budget. Gemini 3.5 Flash‑lite charges $0.30 per 1M input tokens, $0.03 for cached input and $2.5 for every 1M output tokens. By contrast, GPT‑4o mini asks $0.15 for input, $0.075 for cached input and $0.6 for output. In raw terms the OpenAI model costs half as much to feed data but more than three times as much to generate responses, which flips the economics depending on whether your workload is read‑heavy or generation‑heavy.
### Where the dollars actually flow
The table below isolates the token rates that matter for a typical chat‑or‑completion pipeline. Both vendors offer cached‑input discounts that matter when you reuse prompts, but Gemini’s flash tier leans heavily on a low input price while OpenAI leans on a modest output price.
| Vendor | Model | Input | Cached input | Output |
|---|---|---|---|---|
| gemini-3.5-flash-lite | $0.3 per 1M tokens | $0.03 per 1M tokens | $2.5 per 1M tokens | |
| OpenAI | gpt-4o-mini | $0.15 per 1M tokens | $0.075 per 1M tokens | $0.6 per 1M tokens |
If your application is a question‑answer bot that mostly consumes user prompts and only occasionally spits out a short answer, the OpenAI side will feel cheaper because you pay less to ingest data. However, if you are generating long‑form content, code, or multi‑step reasoning, Gemini’s $2.5 output price can dwarf the $0.6 you’d pay to OpenAI, eroding any input savings. The cached‑input discount on Gemini is also shallower, so repeated prompt reuse benefits OpenAI more.
### Who should care about this split
Start‑ups that need a low‑cost entry point for heavy prompt ingestion – for example, analytics dashboards that feed millions of log snippets into a LLM – will likely gravitate toward GPT‑4o mini’s cheaper input rate. Enterprises building content‑generation pipelines, marketing copy farms, or code‑completion services that push large output volumes will find Gemini’s flash tier attractive only if the higher output fee is offset by a massive reduction in prompt engineering overhead or if they can negotiate volume discounts elsewhere. In practice, most mid‑tier users end up balancing the two: they lean on OpenAI for generation‑heavy workloads and switch to Gemini when the workload is input‑dominated.
Pick Gemini 3.5 Flash‑lite if you can keep output tokens low and need the cheapest possible ingestion cost. Choose GPT‑4o mini if your use case leans on generating longer text and you value a more predictable, lower output bill. Pricing verified on 2026-09-10.