Gemini vs GPT: which API actually costs less in 2026?
The two most-compared API families are, at the flagship tier, priced identically on paper — Gemini 3.1 Pro and GPT-5.6 Terra both list $2 input / $12 output. The real cost difference comes from the tiers around them, long-context behavior, and how much of your traffic caches.
Tier-by-tier comparison
| Tier | In / Out | OpenAI | In / Out | |
|---|---|---|---|---|
| Flagship+ | — | — | GPT-5.6 Sol | $5 / $30 |
| Flagship | Gemini 3.1 Pro | $2 / $12 | GPT-5.6 Terra | $2 / $12 |
| Mid | Gemini 3 Flash | $0.50 / $3 | GPT-5.4 mini | $0.75 / $4.50 |
| Small | Gemini 3.5 Flash-Lite | $0.30 / $2.50 | GPT-5.6 Luna | $0.20 / $1.20 |
| Smallest | Gemini 2.5 Flash-Lite | $0.10 / $0.40 | GPT-5.4 nano | $0.20 / $1.25 |
Read down the columns and a pattern appears: Google undercuts OpenAI at the mid and bottom tiers (Gemini 3 Flash beats GPT-5.4 mini; 2.5 Flash-Lite beats nano by 2–3×), while OpenAI's Luna beats Google's newer Flash-Lite in the small tier. At the top, OpenAI sells headroom Google doesn't list a price for (Sol at $5/$30, GPT-5.5 Pro at $30/$180).
A worked example: chatbot workload
Take a support chatbot — 2,000 input + 400 output tokens per request, 100K requests/month, 70% of input cached:
| Model | Monthly cost | Notes |
|---|---|---|
| Gemini 3.1 Pro | ~$628 | identical math to Terra |
| GPT-5.6 Terra | ~$628 | identical math to 3.1 Pro |
| Gemini 3 Flash | ~$157 | 4× cheaper than flagship |
| GPT-5.6 Luna | ~$63 | cheapest capable OpenAI tier |
| Gemini 2.5 Flash-Lite | ~$23 | cheapest overall |
Same conversation, ~27× spread depending on tier choice. Both vendors bill cached input at ~10% of the input rate and both offer a 50% batch discount, so the discounts don't separate them — tier selection does.
The two gotchas
Gemini's long-context surcharge: Pro tiers roughly double above 200K input tokens per request ($4/$18). OpenAI publishes no such surcharge. If you routinely stuff huge prompts, that flips the flagship comparison. OpenAI's small-tier output: Luna's $1.20 output undercuts Gemini's newer Flash-Lite ($2.50) — for output-heavy small workloads OpenAI wins the tier Google usually wins.
Your ratio of input to output tokens decides most of this. Run your own numbers — with the cache and batch sliders — in the calculator, or see the Claude vs GPT comparison and the overall cheapest APIs ranking.
→ What would YOUR workload cost?Enter your token mix once — compare all 28 models side by side, with caching and batch discounts.