Every Google model
Each model on its default cache setting. Select a model for its worked examples, break-even rule and caching rules. Implicit caching figures are a best case, because Google publishes no lifetime and does not guarantee hits.
| Model | Input | Cache read | Output | Min. cacheable | Coding agent | Without | Share saved |
|---|---|---|---|---|---|---|---|
| Gemini 4 Argon | $2.00 | $0.10 | $10.00 | 4,096 | $0.92 | $7.70 | 88% |
| Gemini 3.8 Flash | $0.75 | $0.075 | $3.75 | 4,096 | $0.48 | $2.89 | 83% |
| Gemini 3.6 Flash | $0.75 | $0.075 | $3.75 | 4,096 | $0.48 | $2.89 | 83% |
| Gemini 3.5 Flash-Lite | $0.30 | $0.03 | $2.50 | Not published | $0.22 | $1.18 | 82% |
| Gemini 3.1 Pro Preview | $2.00 | $0.20 | $12.00 | 4,096 | $1.33 | $7.74 | 83% |
| Gemini 3.1 Flash-Lite | $0.25 | $0.025 | $1.50 | Not published | $0.17 | $0.97 | 83% |
| Gemini 2.5 Pro | $1.25 | $0.125 | $10.00 | 2,048 | $0.89 | $4.90 | 82% |
| Gemini 2.5 Flash | $0.30 | $0.03 | $2.50 | 2,048 | $0.22 | $1.18 | 82% |
| Gemini 2.5 Flash-Lite | $0.10 | $0.01 | $0.40 | Not published | $0.062 | $0.38 | 84% |
Coding agent: 40 turns, 20K tokens of starting context, 3K in and 600 out per turn, 1.5 minutes apart, three long breaks of 8 minutes.
How Google caching works
- Implicit caching is on by default for Gemini 2.5 and newer, with no write or storage charge. Google publishes no lifetime and no guarantee of a hit, so implicit figures are a best case.
- Explicit caching, on Gemini 3 and 2.5 models only, caches a fixed prefix with a TTL. Storage is charged per token per hour for the whole TTL.
- Gemini 4 Argon reprocesses the last 2,050 tokens of the previous request when a turn adds fewer than 2,050.
- Google has published price rises for Gemini 4 Argon and the 3.8 and 3.6 Flash models. The simulator switches on their dates.
Sources: pricing page and caching guide, and Google Cloud's caching overview. Each model page lists the rules that apply to it.
Other providers
Anthropic OpenAI All 32 models compared
Try any of these sessions with your own numbers in the simulator. It runs in your browser and shows where the cache goes cold.