Gemini API
Google's Gemini Flash tiers, including a reasoning-budget variant.
3
models in the line
The Gemini models we serve are the Flash tiers — Google's efficiency line rather than its largest models. The interesting one is 3.5 Flash High: the same model as 3.5 Flash but pinned to a high reasoning budget, sold as a separate tier. That makes the choice here partly about how much thinking you want to pay for on identical weights.
At a glance
- From
- $0.9 / 1M tokens
Models in this family
| Model | From | Modes | Open |
|---|---|---|---|
| Gemini 3.6 FlashHigh-efficiency Google model for coding, agentic workflows and web/app development. 1M-token context, text/image/audio/video/file input. | $0.9/ 1M tokens | — | OpenDocs |
| Gemini 3.5 FlashFast general-purpose chat and structured generation through the OpenAI protocol. | $1.05/ 1M tokens | — | OpenDocs |
| Gemini 3.5 Flash (High)Gemini 3.5 Flash pinned to a high reasoning budget — deeper multi-step thinking at a higher token price than the standard tier. | $1.62/ 1M tokens | — | OpenDocs |
Pricing
| Model | Option | Price |
|---|---|---|
| Gemini 3.6 Flash | input | $0.9/ 1M tokens |
| Gemini 3.6 Flash | output | $4.5/ 1M tokens |
| Gemini 3.6 Flash | cacheRead | $0.09/ 1M tokens |
| Gemini 3.5 Flash | input | $1.05/ 1M tokens |
| Gemini 3.5 Flash | output | $6.3/ 1M tokens |
| Gemini 3.5 Flash | cacheRead | $0.1/ 1M tokens |
| Gemini 3.5 Flash (High) | input | $1.62/ 1M tokens |
| Gemini 3.5 Flash (High) | output | $9.72/ 1M tokens |
| Gemini 3.5 Flash (High) | cacheRead | $0.162/ 1M tokens |
Which one to pick
Gemini 3.6 Flash
The newest and cheapest tier, aimed at coding and agentic work, with a 1M-token context and image, audio, video and file input.
Gemini 3.5 Flash
General-purpose chat and structured generation at the standard reasoning budget.
Gemini 3.5 Flash (High)
The same 3.5 Flash pinned to a high reasoning budget — deeper multi-step thinking, at about a 55% premium on input.
Strengths
- 3.6 Flash accepts image, audio, video and file input, which is the broadest input surface in our text lineup.
- A 1M-token context on 3.6 Flash without moving to a frontier-priced model.
- Reasoning budget is selectable through the tier rather than hidden inside the model.
- The newest tier is also the cheapest of the three.
Limitations
- Cache reads are supported but cache writes are not priced here, so caching behaviour differs from the GPT and Claude lines.
- These are Flash tiers only — Google's largest Gemini models are not in this family.
- 3.5 Flash High costs more than 3.5 Flash for the same underlying model, so the premium buys reasoning time, not capability.
The vendor's own description of this line: Google DeepMind — Gemini