Gemini is not one model. It is a ladder of models that trade quality for price, and the gap between rungs is large. Picking the right rung is the single biggest lever you have over cost. This lesson explains the rungs; the chart below shows what each one costs right now.
The three tiers
Pro is the flagship. It reasons more carefully, handles harder coding and analysis, and costs the most. At the time of writing Gemini 2.5 Pro is listed at $1.25 in and $10 out per million tokens, and Gemini 3.1 Pro Preview at $2 in and $12 out.
Flash is the workhorse. It is fast, cheap and good enough for most writing, summarising, extraction and simple code. The Flash line has several generations in the catalogue at once (2.5, 3, 3.5, 3.6, 3.7, 3.8) at prices from $0.30 to $1.50 input. Newer is not always cheaper — Gemini 3.5 Flash is listed above Gemini 3.8 Flash — so check the chart rather than assuming.
Flash Lite is the budget tier for high-volume, low-stakes work: classification, tagging, short rewrites, routing. Gemini 2.5 Flash Lite at $0.10 in and $0.40 out is the cheapest Gemini row we track. Gemini 3.1 Flash Lite and 3.5 Flash Lite sit a little higher.
| Tier | Example model | Input / output | Use it for | Avoid it for |
|---|---|---|---|---|
| Pro | Gemini 3.1 Pro Preview | $2.00 / $12.00 | Hard reasoning, complex code, long multi-step analysis | Bulk summarising, simple Q&A, anything Flash gets right |
| Pro (previous gen) | Gemini 2.5 Pro | $1.25 / $10.00 | Same as above when you want the cheaper Pro | Latency-sensitive chat |
| Flash | Gemini 3.8 Flash | $0.75 / $3.75 | Everyday writing, summaries, extraction, moderate code | Frontier-level reasoning |
| Flash | Gemini 2.5 Flash | $0.30 / $2.50 | Cheapest full Flash; great default for pipelines | Latest features |
| Flash Lite | Gemini 2.5 Flash Lite | $0.10 / $0.40 | Classification, tagging, routing, short rewrites | Nuanced writing, multi-step reasoning |
| Open weights | Gemma 4 31B | $0.09 / $0.34 | Self-hosting, privacy, ultra-cheap bulk text | Long context beyond 262K tokens, multimodal |
Context window: big, but not free
Every current Gemini API model in our catalogue lists a 1,048,576-token context window. That is the headline feature. But context is not free: every token you put in the prompt is billed as input on every call. Pasting a 300-page PDF into a Pro prompt costs real money each time you ask a follow-up question. Lesson 7 explains context caching, which is the fix.
Gemma: the open-weight cousin
Gemma models (ids starting google/gemma-) are open weights. You can download them, run them on your own hardware, or rent them from hosts at very low prices — Gemma 4 31B is listed at $0.09 in and $0.34 out, and some hosts offer free-tier rows. They are text-first, have smaller context windows (131K to 262K tokens in our data) and lag Gemini on hard tasks. For simple, private or enormous-volume text jobs they can beat every Gemini price.
Knowledge check
You need to tag 200,000 support tickets by topic. Which model should you try first?