Gemini 3.5 Flash
$982.50 / month- Input cost
- $307.50
- Output cost
- $675.00
- Per 1,000 calls
- $19.65
- Pricing profile
- standard
Source-tracked model comparison
Gemini 3.5 Flash vs Gemini 3 Flash Preview: what changes in developer-assistant cost?
Compare verified pricing and model limits for moving a Gemini Flash developer workflow to the newer generation. The cost example uses a developer assistant backend workload and does not assume equal model quality.
Direct cost answer: Gemini 3.5 Flash is estimated at $982.50 per month and Gemini 3 Flash Preview at $327.50 for 5,000 input tokens, 1,500 output tokens, and 50,000 monthly requests. Gemini 3 Flash Preview is $655.00 lower under these assumptions. This does not identify a quality winner.
Tracked facts
Pricing and model limits
Prices are USD per 1M tokens under each model's verified default profile.
| Field | Gemini 3.5 Flash | Gemini 3 Flash Preview |
|---|---|---|
| Provider | ||
| API model ID | gemini-3.5-flash | gemini-3-flash-preview |
| Input / 1M | $1.50 | $0.50 |
| Cached input / 1M | $0.15 | $0.05 |
| Output / 1M | $9.00 | $3.00 |
| Context window | 1,048,576 tokens | 1,048,576 tokens |
| Maximum output | 65,536 tokens | 65,536 tokens |
| Accepted input | text | text, image, audio |
Example workload
Developer assistant backend cost scenario
5,000 input and 1,500 output tokens per request, 50,000 monthly requests, and 20% cached input.
Gemini 3 Flash Preview
$327.50 / month- Input cost
- $102.50
- Output cost
- $225.00
- Per 1,000 calls
- $6.55
- Pricing profile
- standard
Cost result: Gemini 3 Flash Preview is $655.00 lower per month for these assumptions. This is a price comparison, not a model-quality ranking.
Questions to answer before choosing
- Which developer tasks improve in a representative test?
- Does the existing Gemini integration need changes?
- How do output length and retries change monthly cost?
What this estimate leaves out
The calculation does not compare coding quality and excludes indexing, execution sandboxes, and repository storage.
Latency, reliability, output quality, retries, regional processing, and provider-specific tool charges can change the practical decision.