Google released Gemini 3.6 Flash on 21 July 2026. At $0.75 per million input tokens and $3.75 per million output tokens, it is half the launch price of Gemini 3.5 Flash, which started at $1.50 and $9.00.
For high-volume multimodal work, that is the whole story.
For the three-vendor picture, see the full AI model comparison for 2026.
Pricing
| Tier | Input per 1M | Output per 1M |
|---|---|---|
| Gemini 3.6 Flash, standard | $0.75 | $3.75 |
| Gemini 3.6 Flash, batch | $0.375 | $1.875 |
| Gemini 3.5 Flash, launch price | $1.50 | $9.00 |
These are list prices per million tokens, checked on 3 November 2026. Confirm the rate for the tier you will use.
A note on the June numbers: the original comparison listed Gemini 3.5 Flash at about $0.30 and $2.50. Those figures matched Flash-Lite, not Flash. The correction is documented on the comparison page.
What a workload costs
Take 100 million input tokens and 20 million output tokens:
| Setup | Monthly cost |
|---|---|
| Gemini 3.6 Flash, standard | 100 × $0.75 + 20 × $3.75 = $150 |
| Gemini 3.6 Flash, batch | 100 × $0.375 + 20 × $1.875 = $75 |
| GPT-6.1 Sol, direct | 100 × $1 + 20 × $5 = $200 |
| Claude Fable 5.1 | 100 × $10 + 20 × $50 = $2,000 |
Gemini 3.6 Flash is cheaper than Sol and more than 13 times cheaper than Fable 5.1 on the same volume. For a pipeline that processes millions of documents, that difference decides whether the project is viable at all.
What changed from 3.5 Flash
| Gemini 3.5 Flash | Gemini 3.6 Flash | |
|---|---|---|
| Launch price | $1.50 / $9.00 | $0.75 / $3.75 |
| Multimodal | Text, image, audio, video | Same |
| Speed | Fast | Fast |
| Computer use | Built-in | Built-in |
| Position | Frontier intelligence at Flash speed | Cheaper volume model |
The first change is price: half the launch rate. The second change is strategic. Computer use is no longer exclusive to Gemini, because GPT-6.1 Terra now includes it. Gemini’s advantage is now cost and native multimodality, not a unique feature.
Where Gemini 3.6 Flash fits
- High-volume extraction from documents, invoices, screenshots and forms.
- Video and audio analysis where one model needs to handle every media type.
- Batch jobs such as overnight classification, backfills and bulk summarisation.
- Real-time features where latency matters more than the last point of reasoning.
- Cost-sensitive agents with simple tool loops and clear success criteria.
Where it does not fit
- The hardest coding tasks. Claude Fable 5.1 remains the stronger choice for demanding code work.
- Long-horizon agent work where one wrong step is expensive to unwind.
- High-stakes reasoning where you would pay for GPT-6.1 Sol and accept the cost.
Use batch where you can
Batch pricing is half the standard rate. Any job that does not need an answer in seconds is a candidate: overnight re-indexing, monthly reporting, bulk content classification and backfills.
Two rules make batch easy to adopt:
- Separate interactive traffic from background traffic in your code.
- Send background work to batch by default, and move it back to standard only when a user is waiting.
Should you use it for coding?
Gemini 3.6 Flash is competitive on general coding and strong on fast iteration, but it is not the model to choose for the hardest repository-scale work. Use it for code explanation, test generation, small refactors and review assistance. Use Fable 5.1 for migrations and complex agentic coding.
A simple decision rule
| If you need… | Use |
|---|---|
| High volume at the lowest cost | Gemini 3.6 Flash |
| Native audio or video processing | Gemini 3.6 Flash |
| The strongest coding quality | Claude Fable 5.1 |
| Flagship reasoning for hard tasks | GPT-6.1 Sol |
| Computer use on a budget | GPT-6.1 Terra or Gemini 3.6 Flash |
How to evaluate it
- Take 50 real inputs from production, including the ugly ones.
- Compare output quality against your current model, scored blind.
- Measure cost per successful task, not per token.
- Test p95 latency on your own stack.
- Check multilingual and multimodal cases if your users need them.
FAQ
Is Gemini 3.6 Flash cheaper than 3.5 Flash?
Yes. It launched at $0.75 input and $3.75 output per million tokens, half the $1.50 and $9.00 launch price of Gemini 3.5 Flash.
Is Gemini 3.6 Flash good for coding?
It is competitive for general coding, explanation, tests and small refactors. For the hardest coding and long agent runs, Claude Fable 5.1 is the stronger choice.
Does it handle video and audio?
Yes. Gemini is natively multimodal across text, image, audio and video in a single model.
Should I use the batch tier?
If the job does not need an answer in seconds, yes. Batch is half price and suits overnight and bulk workloads well.
Next step
If you are building an AI feature and need help estimating the cost, our AI app development cost guide covers the build side. For a short evaluation on your own data, contact 24Bit System.
Related reading: GPT-6.1 Sol vs Terra and Claude Fable 5.1.
