Google models on One API
20 models we could call on 22 November 2026, from $0.0675 per million input tokens.
Google in numbers
- Models we could call
- 20
- our probe of 22 November 2026 · measured by us
- Median time to first token
- 1,210 ms
- our probe of 22 November 2026 · measured by us
Every Google model
| Model | Input, $ per 1M | Output, $ per 1M | Context, tokens | First token | Tokens a second |
|---|---|---|---|---|---|
| gemma-4-26b-a4b-it | $0.0675 | $0.225 | 262,144 | 982 ms | not measured |
| gemini-2.5-flash-lite | $0.1 | $0.4 | 1,048,576 | 531 ms | 90.6 |
| gemini-embedding-001 | $0.15 | n/a | 2,048 | 362 ms | n/a |
| gemma-4-31b-it | $0.2 | $0.4 | 262,144 | 1,103 ms | not measured |
| gemini-embedding-2 | $0.2 | n/a | 8,192 | 391 ms | n/a |
| gemini-3.1-flash-lite | $0.25 | $1.5 | 1,000,000 | 858 ms | 739.6 |
| gemini-3.1-flash-lite-image | $0.25 | $1.5 | 65,536 | 963 ms | 476.3 |
| gemini-3.1-flash-lite-preview | $0.25 | $1.5 | 1,048,576 | 793 ms | 758.8 |
| gemini-2.5-flash | $0.3 | $2.5 | 1,000,000 | 1,191 ms | not measured |
| gemini-2.5-flash-image | $0.3 | $2.5 | 32,768 | 724 ms | 32.6 |
| gemini-3.5-flash-lite | $0.3 | $2.5 | 1,048,576 | 850 ms | 715.2 |
| gemini-3-flash-preview | $0.5 | $3 | 1,000,000 | 1,377 ms | not measured |
| gemini-3.1-flash-image | $0.5 | $3 | 131,072 | 1,228 ms | 302.2 |
| gemini-3.6-flash | $0.75 | $3.75 | 1,048,576 | 1,713 ms | not measured |
| gemini-3.8-flash | $0.825 | $4.125 | 1,048,576 | 2,820 ms | not measured |
| gemini-robotics-er-2-preview | $1 | $5 | 131,072 | 1,282 ms | not measured |
| gemini-2.5-pro | $1.25 | $10 | 2,000,000 | 2,228 ms | not measured |
| gemini-3.5-flash | $1.5 | $9 | 1,000,000 | 2,069 ms | not measured |
| gemini-3-pro-image | $2 | $12 | 65,536 | 3,980 ms | 239.9 |
| gemini-3.1-pro-preview | $2 | $12 | 1,048,576 | 4,342 ms | not measured |
Other providers
How this page is made
Prices, context sizes and abilities are read from One API's public price feed and its model endpoints each time this page is built. The time to the first token, the output speed and the price of a call are measured by us: 3 calls for each model through a One API key on 22 November 2026, entering at not recorded. A model that did not answer has no page. How we measure.
Data checked · Last changed