Twenty-two models from twelve companies, in one place, with the right one picked for each question instead of whichever company you happened to subscribe to. From $4.99 a month.
fast & cheapeverydaylong documentscodedeep reasoning
What people actually use it for
Switchboard is a general assistant first. It happens to be unusually good at building things, which matters if that’s what you came for — and costs you nothing if it isn’t.
Every day
Draft the awkward email. Plan the trip. Explain the letter from the bank. Summarise the forty-page PDF. Settle the argument. Write the toast.
At work
Quotes and proposals, meeting notes into actions, a spreadsheet that adds up, research with the sources attached. Export to Word, Excel or PDF when it needs to leave the chat.
If you build
Write a website and look at it before you download it. Connect your own GitHub, Supabase or Linear and work on your real systems, read-only until you say otherwise.
The argument
The cheap tiers elsewhere are cheap because they hold the best models back: you get the quick one, and the one that actually thinks costs twenty dollars. Switchboard Lite reaches every model we carry. A smaller monthly amount of them — not a smaller menu.
Anthropic’s cheapest plan is $20. Ours reaches Claude Opus 5 at $4.99.
| Plan | Company | Per month |
|---|---|---|
| Lite | Switchboard | $4.99 |
| Google AI Plus | $7.99 | |
| ChatGPT Go | OpenAI | $8.00 |
| Plus | Switchboard | $9.99 |
| Pro | Switchboard | $19.99 |
| Google AI Pro | $19.99 | |
| ChatGPT Plus | OpenAI | $20.00 |
| Claude Pro | Anthropic | $20.00 |
| Builder | Switchboard | $49.99 |
Advertised US consumer prices, checked September 2026. Above $50 sit ChatGPT Pro, Claude Max and Google AI Ultra — we don’t compete there. Check for yourself at OpenAI, Anthropic and Google.
Side by side
Two fights, and they’re different ones. At the cheap end it’s about what you can reach. At twenty dollars it’s about how many companies you get.
| Switchboard Lite — $4.99 | ChatGPT Go — $8.00 | Google AI Plus — $7.99 | Claude | |
|---|---|---|---|---|
| The best reasoning models | All of them | Thinking costs $20 | 30 prompts a day | No plan under $20 |
| Models from other companies | 22 across 12 | OpenAI only | Google only | Anthropic only |
| Largest context you can reach | Up to 1M | More than free | 128K | — |
| Connect your own tools | Included | — | — | — |
| Switchboard | ChatGPT Plus | Claude Pro | Google AI Pro | |
|---|---|---|---|---|
| Entry price a month | $4.99 | $20.00 | $20.00 | $19.99 |
| Models from more than one company | Yes | OpenAI only | Anthropic only | Google only |
| Picks the right model for each message | Across companies | Within theirs | Within theirs | Within theirs |
| Shows which model answered, and what it cost | Every answer | No | No | No |
| Spending caps so a bill can’t run away | Built in | Flat fee, then limits | Flat fee, then limits | Flat fee, then limits |
| Connect your own tools by signing in | From $4.99 | From $20.00 | From $20.00 | Workspace plans |
| Voice conversation | Hands-free mode | Yes | No | Yes |
Pictures
Both of these were made by the product, not for the page. The model and the price under each are the ones its own ledger recorded.
“Create an image of a red delivery van parked outside a body shop”

“Crop this to just the red van and put it on a plain white background”

The second one is the half people don’t expect. Photograph a panel, say what to take out, get it back. There is a dearer tier for when it has to be right — Qwen Image 2.1 Pro, at $0.046 — and pictures are counted in pictures rather than tokens, so what you have left is a number you can read.
Where they beat us
A comparison table that says we win every row is a comparison table nobody believes.
The broadest toolkit in one place, and the most polished voice mode. If you talk to it while driving, or you live inside custom GPTs, stay there.
Exceptional on very long documents and sustained writing, with agentic coding tools of its own. If one enormous manuscript is your whole job, it’s worth the twenty.
If your work already lives in Gmail, Docs and Drive, having the assistant inside them is worth more than anything we can offer from outside.
Switchboard is for everyone else: people who want several companies’ models without paying several subscriptions, who’d rather spend five dollars than twenty, and who want to see where the money went.
How it works
In plain language. No model names, no dropdown, no wondering whether this one is worth the expensive option.
A quick question goes to a fast, cheap model. A hard one goes to a strong one. If a model fails, the next takes it before you’ve read a word.
Every answer shows which model wrote it and what it cost your allowance. Your plan has a ceiling, so a heavy month can’t become a surprise.
Plans
Every plan includes the whole product. What changes is how much you can use and how many of your own tools you can connect.
Lite
$4.99/mo
Most people
Plus
$9.99/mo
Pro
$19.99/mo
Builder
$49.99/mo
Before you ask
No, and most of what people do here isn’t technical at all. It writes your emails, reads your documents, plans your week and answers your questions like any good assistant. The building tools sit out of the way until you want them.
It’s a switchboard, and that’s the point — we don’t train models, we choose between them. What we built is the part around them: routing each message to a sensible model, keeping the cost of each answer visible, capping your spend, connecting your own tools safely, and turning an answer into a document or a working website. Nobody is going to out-train OpenAI. Getting you the right model for five dollars is a different job.
Twenty-two, from twelve companies — OpenAI, Anthropic, Google, Meta, xAI, DeepSeek, Mistral, Alibaba and others — chosen per message and updated as better ones arrive. Claude Opus 5, GPT-5.6 Terra, Grok 4.6 and Kimi K3 are all reachable on the $4.99 plan. You never pick one, but you always see which one answered, and one click sends the same question to a stronger model if the first disappoints.
Yes, and it can change ones you already have — attach a photo and say what to do, like “take the skip out of the background” or “put this on plain white for the quote”. Every plan from $4.99 includes a monthly allowance of both, counted in pictures rather than tokens, so you always know what you have left. ChatGPT and Gemini make pictures too, at $8 and $7.99; Claude doesn’t make them at all. What’s different here is the same thing as everywhere else — you see which model drew it and what it cost.
You’re told before you get there, not after. You can wait for the window to reset, add a top-up, or move up a plan. Nothing auto-charges and nothing silently degrades.
No. Your conversations are yours, and every provider we send them to is configured to refuse data collection. Connected tools stay read-only until you turn changes on yourself, every tool run is written to your activity log, and access tokens are encrypted and never shown again. The one exception is above, not in here: if every paid model is down, the free questions on this page can fall back to a free-tier model that may train on them — you are told at the time it happens, and it is never used for a signed-in account.
Any time, from your own settings, without emailing anybody.