Flux vs going direct to each provider
Going direct means an account with each AI lab, and another for each tool your agent uses. You get each lab's full feature set at its own price, with nobody in the middle. Flux puts every model on one key and one bill, and picks a model for each job.
Checked 28 September 2026.
Who each is for
Teams settled on one lab, with the time to manage its account, limits and features themselves.
Anyone whose agent should use the best model for each job without an account at every lab.
Side by side
| Capability | Flux | Going direct |
|---|---|---|
| Every major model on one key | Yes | No. One per vendor |
| Picks the right model for each job | Yes | No |
| Live web search and page reading | Partly. Page reading now | Partly. Some |
| Voices, speaking and listening | Partly. Transcription now | Partly. Per vendor |
| Actions in apps | Partly. Coming to your key | No |
| Memory across every model | Partly. Shipping | Partly. Per vendor |
- Models
- Flux90+ models from 12 labs, on one key.
- Going directOne lab's models per account, and a new account for each lab.
- How the model is chosen
- FluxSend flux-auto and Flux picks among vetted models for each kind of task, cost-aware. Or name any model in the catalogue yourself.
- Going directYou choose every call.
- Pricing
- FluxOne flat rate per lane: Fast, Standard and Reasoning. flux-auto bills at the Standard rate, or the Fast rate when it sends your request to the Fast lane and a Fast model answers. Never more than Standard. A model you name bills at the rate on the pricing page.
- Going directEach lab's own rates, billed separately by each lab.
- Tools for your agent
- FluxPage reading, transcription and images on the same key.
- Going directA separate account and bill for each tool vendor.
Where each one wins
Where going direct wins
- One provider and one contract.
- Each lab's newest features and full control of prompt caching.
- Nobody in the middle, and no gateway margin.
Where Flux wins
- Every model on one key and one bill.
- A model picked for each job.
- Page reading, transcription and images on the same key.
Which to choose
Go direct if you use one model from one lab and rely on its prompt caching or newest features.
Choose Flux if your agent uses more than one model, or needs to read the web or transcribe audio.
Questions
Is going direct cheaper?
For one model from one lab, usually yes: you pay the lab's own rate and nothing else. Flux bills one flat rate per lane, and flux-auto bills at the Standard rate, or the Fast rate when it sends your request to the Fast lane and a Fast model answers.
Do I lose prompt caching with Flux?
Not entirely. Flux adds prompt caching automatically for some labs. Going direct gives you full control of it, which matters for long, repeated prompts.
Can I still use one lab through Flux?
Yes. Name the model you want and Flux sends every call there, at the rate on the pricing page.
Sources
This page compares Flux with keeping your own account at each lab. It names no vendor, so there are no third-party figures to cite. Flux facts come from our own catalogue and pricing.
We re-check these facts every quarter. If something here is out of date, tell us and we will fix it.
Try Flux with the agent you already use.
One key, any agent. Point the tool you already use at Flux and keep everything else.