TypingMind and ChatGPT are not the same kind of purchase, and that is the whole comparison: TypingMind is a browser front end licensed once — $39, $79 or $99 — that you then feed with your own provider API keys, while ChatGPT is a subscription — $20 per month for Plus — that bundles the models with a first-party product no API sells. Which is cheaper depends on your monthly token volume and your model choice, and the crossover falls inside the range ordinary people actually use. The comparison only works if you keep TypingMind's licence beside API token spend and ChatGPT's subscription beside what it includes — crossing those two is the most common way this question gets answered wrongly.
Two disambiguations first, because both cost people money. typingmind.com/buy sells a one-time individual licence; custom.typingmind.com/pricing sells TypingMind Teams, a separate product on a recurring subscription with a different architecture; a third thing shares the name, the TypingMind Custom API, which administers a Teams workspace and is not a way to buy model tokens. And ChatGPT Team was renamed ChatGPT Business on August 29, 2025 — OpenAI's help centre says the rename itself changed no features, pricing or usage limits — so any source still saying "ChatGPT for Teams", including TypingMind's own comparison page, is describing the old name.
Which one you should pick
Decide on the job before the feature grid, because the two products fail in different places.
| If this is your situation | Pick | What you give up |
|---|---|---|
| You want one bill, zero configuration, and you use ChatGPT's own tooling | ChatGPT Plus at $20 per month | Per-plan context caps, and no documented way to point it at another model provider |
| You want several vendors' models behind one interface and no recurring software fee | A bring-your-own-key client; TypingMind is one | You assemble the features yourself, and TypingMind's tier ladder gates some of them two levels up |
| Chats must stay on your own machine unless you say otherwise | TypingMind, whose docs say chats are saved locally in the browser by default with opt-in sync | Backups and device-to-device sync become your job, or a paid add-on |
| You need code interpreter / advanced data analysis | ChatGPT | TypingMind's own Teams-vs-ChatGPT comparison page marks that row "Coming soon" against a checkmark for ChatGPT |
| A team needs one admin-held key rather than per-person keys | TypingMind Teams or ChatGPT Business | Both are recurring; ChatGPT Business requires at least two paid seats |
| You are leaving ChatGPT and want your history to come with you | TypingMind, which documents importing a ChatGPT Plus export | The reverse trip has no documented importer in any client checked, so plan on re-entering configuration by hand |
| Your model access comes from a gateway that sends no browser CORS headers | A desktop or self-hosted client — not TypingMind's browser app | TypingMind's documented install is a browser app, so such an endpoint is blocked before the request leaves the page (measured below) |
What each one actually costs
TypingMind's individual licence is one-time. Its buy page says "Buy once, use forever", "One-time, no renewal" and offers a 14-day money-back guarantee, with the footnote that matters: "Price does not include API costs and additional cloud storage".
| TypingMind individual licence | One-time price | What the tier adds |
|---|---|---|
| Standard | $39 | Remove ads and popups, basic chat features, AI agents, share chats and plugins, voice input |
| Extended | $79 | Web search, vision and images, image generation, document upload, voice output |
| Premium | $99, shown with a "50% OFF" badge against a struck-through $198 | Multi-model parallel chats, projects and folders, artifacts and canvas, unlimited plugins and MCPs, free updates |
| Bulk licence | $395, shown with a "50% off" badge against a struck-through $790 | "A shared License Key for up to 10 users on 50 different devices" |
Read from the buy page on September 21, 2026. Because Premium and the bulk licence are both shown at half price against struck-through figures, with no end date printed anywhere on the page, treat $99 as the price shown today rather than a documented permanent list price. A recurring line item can still sit behind "one-time", if you opt into cloud sync: the cloud-storage documentation gives 50MB free, "typically enough for 5000 chats", then offers 1GB and 5GB plans payable monthly or annually and prints no figures for either. Those prices are visible only inside the app, so this page will not guess them.
ChatGPT is the opposite shape — recurring, with the models included.
| ChatGPT plan | Published price | Notes worth knowing |
|---|---|---|
| Free | $0 | Everyday text chat, with limited uploads, image generation, voice and deep research |
| Go | Not stated here — see the note below | The tier OpenAI says may include ads; monthly only; API usage not included |
| Plus | $20 per month, billed monthly | "Currently, we do not support annual billing"; "API usage is separate and billed independently" |
| Pro | Not stated here — see the note below | Two tiers, named "Pro $100" and "Pro $200 (Pro 20X)"; new sign-ups and upgrades to the $200 tier have been paused since September 10, 2026 |
| Business, Standard seat | $25 per user per month, or $20 billed annually | At least two paid seats; from August 24, 2026 a maximum of 200 paid seats per subscription |
| Business, Premium seat | $125 per user per month, or $100 billed annually | Five times a Standard seat's usage, and not subject to the five-hour usage limit |
| Enterprise | Contact sales | No published figure |
Plus and Business figures come from OpenAI's own English help centre, checked September 21, 2026; OpenAI adds that prices are in USD and "may vary by country and currency". Two prices are deliberately blank. Go: OpenAI's January 16, 2026 announcement gives $8 per month as US pricing and footnotes that Go is localised in some markets, while the live pricing page reachable from this network geo-redirects and renders US$7, and the Go help article prints no price at all. Pro: the English help centre names two tiers without stating a figure, and the same pricing page shows "starting at US$120". Three numbers with no reconciliation is not a price, so confirm both at OpenAI's own checkout.
One more thing that makes the ChatGPT side less flat than it looks. OpenAI documents usage credits for Free, Go, Plus and Pro that "let you continue using eligible features after reaching your plan's included limits" — Codex, ChatGPT Work, Word, Excel and PowerPoint, not ordinary chat. They are "valid for 12 months from purchase", expire unused, are non-refundable, and OpenAI states they are not API credits. So both sides of this comparison have a variable second line item; they simply sit in different places.
TypingMind's own API-cost claim is the trap
The buy page's FAQ still answers the cost question like this, verbatim on September 21, 2026: "The ChatGPT API costs soooooo little you won't even notice it. Typical chat conversation is about 1000 characters, which costs ~$0.001. That's less than a cent. You can have thousands of chats and it will only costs you like $1." That is a GPT-3.5-era anchor, and budgeting from it is how people end up surprised. Against OpenAI's current list, one thousand characters is on the order of 250 tokens — a rule of thumb, not a measurement — and 250 output tokens on GPT-6 Astra at $50.00 per 1M is about $0.0125 on its own, before any input tokens, before the conversation history every later turn resends, and before the reasoning tokens you never see.
Here is the same question answered with volumes instead. Assume a light month of 375,000 input and 45,000 output tokens and a heavy month of 1,500,000 input and 180,000 output — roughly 15 versus 60 substantial conversations, an assumption for comparison rather than a measurement. These are illustrative token arithmetic, not measured bills and not a ceiling.
| Model, at OpenAI's own API rates | Input / output per 1M | Light month | Heavy month |
|---|---|---|---|
| GPT-6 Astra | $10.00 / $50.00 | $6.00 | $24.00 |
| GPT-5.6 Terra | $2.00 / $12.00 | $1.29 | $5.16 |
| GPT-5.6 Luna | $0.20 / $1.20 | $0.13 | $0.52 |
Rates read from OpenAI's API pricing page on September 21, 2026. The reading that decides this comparison: a heavy month on the flagship is $24.00 in tokens alone, already above ChatGPT Plus at $20 per month before the licence is counted — while the same month on a mid-tier model is $5.16. Over a first year with a $99 Premium licence that is $387.00 against $160.92, versus $240.00 for twelve months of Plus. Model choice moves this comparison further than the software does.
A gateway changes the token column rather than the structure. At live Kunavo catalog rates the same assumed months look like this, across two vendors because a key-based client is not confined to one.
| Model on Kunavo | Input / output per 1M | Light month | Heavy month |
|---|---|---|---|
| GPT-6 Astra | $4.00 / $20.00 | $2.40 | $9.60 |
| GPT-5.6 Terra | $0.70 / $4.20 | $0.45 | $1.81 |
| GPT-5.6 Luna | $0.07 / $0.42 | $0.05 | $0.18 |
| Claude Sonnet 4.6 | $1.20 / $6.00 | $0.72 | $2.88 |
On these assumptions a heavy flagship month is $9.60 in tokens, against $24.00 for the same month at OpenAI's own rate, and twelve such months is $115.20 versus $240.00 of Plus. No licence is added to that column on purpose: this route needs a client that can actually reach the gateway, and the section below measures why TypingMind is not one. Read the figures as arithmetic on assumed volumes, and note that cheapest listed price and lowest cost to finish the task are different claims — a weaker model you re-prompt twice can cost more than a stronger one answering once. Kunavo's catalog amount is a billing floor rather than a cap: when the upstream reports its charge, the bill is the greater of catalog cost and upstream cost times the applicable markup. Cache charges and external tools sit outside these examples. The minimum top-up is $10 in prepaid credit — a funding minimum, not a task fee or a subscription — and the balance does not expire, which is the opposite of how ChatGPT's usage credits behave. See billing details.
What the subscription buys that a key cannot
This is where a bring-your-own-key page has to stop selling. ChatGPT's paid plans list a long set of first-party capabilities that are product features of the ChatGPT application, not endpoints: ChatGPT Work, Codex, scheduled tasks, shared projects, sites, a built-in browser, workspace agents, company knowledge, connectors to internal tools, the Word, Excel, PowerPoint and Google Sheets extensions, record mode, discovering and publishing custom GPTs, study mode, voice with video, and memory of past conversations. None of them is sold through a model API, so a front end gets none of them from the endpoint it points at. That is not the same as saying nobody else can build the idea — TypingMind ships its own projects, folders and agents, as the licence table above shows — but the ChatGPT versions, and the ecosystem around them, stay inside ChatGPT. TypingMind's own Teams-vs-ChatGPT comparison page concedes one capability outright: the code-interpreter row is a checkmark for ChatGPT and "Coming soon" for TypingMind.
The advantage runs the other way on context. The ChatGPT plan comparison table caps the total context window by plan — 27K, 54K, 54K and 128K for the instant model on Free, Go, Plus and Pro, and 256K on Go and Plus against 400K on Pro for the reasoning model, with a footnote that the space usable for your own input is smaller still because system instructions and memory consume part of it. The API tier is not subject to those plan caps: Kunavo's catalog declares a 1,050,000-token context window on GPT-6 Astra. Read from the pricing comparison table on September 21, 2026 — this network reaches it only in its zh-Hans-CN rendering, so the figures are that rendering's; the USD amounts elsewhere on the same page matched OpenAI's English help centre exactly, which is evidence rather than proof that the table is not localised in substance.
Note also that TypingMind's documentation is stale on OpenAI's lineup — its Teams-vs-ChatGPT comparison page still says "GPT-4o Usage Limit" and prices ChatGPT at "$30/month/user", and its OpenAI provider page still tells you that you can now use "GPT-5.4, Codex, GPT-5". Take OpenAI's model names and prices from OpenAI. The product itself is current: its live changelog added GPT-6 Astra on September 7, 2026 and its newest entry is dated September 21, 2026.
Where a third-party endpoint can and cannot go
ChatGPT first, because the answer is short: OpenAI documents no way to point ChatGPT at a third-party or self-hosted model endpoint on any plan. Connectors, custom GPT Actions and the Apps SDK let ChatGPT call external tools; they do not swap the model. That is an absence of a documented feature rather than a prohibition. The documented corollary is the budget one — Plus states "API usage is separate and billed independently" and Go says API usage "is not included and billed independently", so a ChatGPT subscription buys nothing on the API side.
TypingMind does take a custom endpoint, with two conditions. The field wants the complete route, not a base URL — its Proxy documentation prints defaults of https://api.openai.com/v1/chat/completions and https://api.anthropic.com/v1/messages, and its Add Custom Model flow has you type the auth header by hand, name Authorization, value Bearer plus your key. Its OpenAI page adds that a customised endpoint "must be 100% compatible with OpenAI's Chat Completion endpoint", and its custom-model walkthroughs tell you to "enter the exact endpoint". Both documented defaults are complete routes, so a bare base URL is not what that field expects.
The second condition is the one that decides which gateways work at all. TypingMind is a browser app — its install page documents a Progressive Web App on every platform, "no app store, no download required", and documents no native build — so on the routes that page covers, every model call is a cross-origin request from the typingmind.com origin, or from whatever origin a self-host copy is served from. TypingMind concedes the constraint in its own MCP documentation, which says "most MCP servers don't allow direct connection between TypingMind app running in the browsers to their server" and that its MCP Connector "is only needed in the Personal version".
| Preflight from the typingmind.com origin | Result on September 21, 2026 | Reachable from TypingMind? |
|---|---|---|
api.openai.com/v1/chat/completions | 200, access-control-allow-origin: https://www.typingmind.com | Yes |
api.anthropic.com/v1/messages | 200, access-control-allow-origin: *, once the request declares anthropic-dangerous-direct-browser-access | Yes |
api.kunavo.com/v1/chat/completions and /v1/messages | 204, with no access-control-allow-origin header in the response | No — blocked at preflight |
Those preflights were run from a terminal for this page on September 21, 2026. So Kunavo cannot be configured as a model endpoint in the TypingMind browser app or its installed PWA today, and self-hosting TypingMind does not change it, because the request still starts in a web page. TypingMind is also listed on Setapp as a Mac app; nothing here tested that build, so this conclusion is stated for the routes TypingMind's own install page documents rather than for every way it is distributed. TypingMind Teams is architecturally different — its pricing FAQ answers "Do my users see my OpenAI API key?" with "No. Your users will not see your API key. Requests to the OpenAI API are sent from our server", and server-originated calls are not subject to browser CORS — but that statement is about the OpenAI provider, nothing here tested whether Teams routes arbitrary custom endpoints the same way, and Teams is a recurring subscription either way. Name the billing toggle when you quote its price, because the page shows two sets: Starter, Growth and Professional at $99, $199 and $299 per month with Monthly Billing selected, and $83, $166 and $249 per month on the Annual toggle, which carries a "2 MONTHS FREE" badge and is the set a plain fetch of the page returned on September 21, 2026. Professional is the one to read twice — on both toggles it is shown as a half-price offer against a struck-through figure, $599 monthly and $499 annual. Starter includes five seats and extra seats are $8 per seat per month, and a footnote under the plan grid says API cost is not included. Confirm which toggle you are on at checkout. Teams is not a route this page will recommend for reaching a gateway.
If a gateway is what you want, use a client that issues the call from its own process. Kunavo publishes configuration pages for Chatbox, Cherry Studio, Open WebUI and LobeHub. A configuration page is a reference, not a compatibility test — none of these clients, and neither product on this page, has been runtime-tested against Kunavo's endpoint — so keep a working provider available while you try one, run a bounded session, and read what your account actually recorded. Create a Kunavo account when you want to fund a key.
One boundary before you design a workflow around either product: Kunavo serves no embedding, speech-to-text or text-to-speech model, so retrieval over your documents and any voice feature stay with the client or another provider regardless of which gateway answers the chat calls. For the endpoint conventions in general see OpenAI-compatible API; for the wider client field, TypingMind alternatives; for OpenAI's token rates in detail, GPT API pricing; and if the comparison you actually want is the coding one, Codex vs ChatGPT.
FAQ
Is TypingMind better than ChatGPT?
They are different purchases, so the honest answer is "for which job". TypingMind is a browser front end you licence once — $39, $79 or $99 one-time on its buy page, checked September 21, 2026 — and then supply with your own provider API keys, which is how it reaches models from several vendors behind one interface. ChatGPT is a subscription, $20 per month for Plus per OpenAI's own help centre, that bundles the models with a first-party product: Codex, ChatGPT Work, the Word, Excel and PowerPoint extensions, connectors to internal tools, scheduled tasks, custom GPTs, study mode and memory. None of those is sold through a model API, so a bring-your-own-key client gets none of them from the endpoint it points at — the generic ideas it can build itself, and TypingMind builds several, but the ChatGPT implementations stay inside ChatGPT. TypingMind wins when you want many vendors' models, local-first chat storage and no recurring software fee; ChatGPT wins when you want the product features and one bill.
Is TypingMind cheaper than ChatGPT Plus?
Only at low-to-moderate volume or on a cheaper model, and the crossover is real rather than theoretical. ChatGPT Plus is $20 per month billed monthly with no annual option (OpenAI help centre, checked September 21, 2026). TypingMind is a one-time licence plus whatever your API keys spend. Using OpenAI's own published API list prices on September 21, 2026 — GPT-6 Astra at $10.00 input and $50.00 output per 1M tokens — an assumed heavy month of 1,500,000 input and 180,000 output tokens is $24.00 in tokens alone, above the Plus subscription before the licence is counted. The same assumed month on GPT-5.6 Terra at $2.00 and $12.00 per 1M is $5.16. So the flagship model makes the bring-your-own-key route more expensive and a mid-tier model makes it cheaper, on identical usage. These are illustrative token arithmetic on assumed volumes, not measured bills.
Is TypingMind open source?
No, and the GitHub repository is why people think otherwise. github.com/TypingMind/typingmind is public, not archived, has a src/ directory and recent commits, and GitHub reports its licence as NOASSERTION. The LICENSE.md in that repository says verbatim: "NOTE: This is NOT an open-source software." and "The source code here is compiled and is intended only for self-host users", adding "You do not have permission to modify or redistribute the compiled code." Checked through the GitHub API on September 21, 2026. So TypingMind cannot be compared with AGPL or Apache-licensed clients on licence grounds — the self-host package is a compiled build under proprietary terms, not source you may fork.
Can TypingMind use a custom API endpoint or a gateway?
Sometimes, and two conditions decide it. First, TypingMind's endpoint field takes the complete route rather than a base URL — its documented defaults are https://api.openai.com/v1/chat/completions and https://api.anthropic.com/v1/messages, and its Add Custom Model flow has you type the Authorization header by hand as "Bearer your_api_key". Its OpenAI page also states a customised endpoint "must be 100% compatible with OpenAI's Chat Completion endpoint". Second, because TypingMind runs in a browser, the endpoint must return CORS headers permitting the typingmind.com origin, or the browser blocks the call at preflight before the server sees it. Checked from a terminal on September 21, 2026: api.openai.com reflects that origin back, api.anthropic.com returns a wildcard once a request declares anthropic-dangerous-direct-browser-access, and api.kunavo.com returns 204 with no access-control-allow-origin at all. Kunavo therefore cannot be configured in TypingMind's browser app or its installed PWA today, and self-hosting TypingMind does not change it because the request still originates in a web page. TypingMind also has a Setapp Mac listing that nothing here tested, so read this as covering the routes its install documentation describes.
Can ChatGPT be pointed at a different model provider?
OpenAI documents no mechanism for it on any plan — Free, Go, Plus, Pro, Business or Enterprise. Every model row on the ChatGPT plan comparison table is an OpenAI model, and connectors, custom GPT Actions and the Apps SDK let ChatGPT call external tools rather than swap the model behind the conversation. That is an absence of a documented feature rather than a vendor prohibition, so read it as "no published route" rather than "forbidden". The documented corollary is the one that affects budgets: OpenAI states that API usage is separate and billed independently on both the Plus and the Go plans, so a ChatGPT subscription buys nothing on the API side, and API credit is exactly what a bring-your-own-key client such as TypingMind spends.
Is a ChatGPT subscription still a flat monthly fee?
Not entirely, as of the credits system OpenAI documents for Free, Go, Plus and Pro. Its help centre says credits "let you continue using eligible features after reaching your plan's included limits", naming Codex, ChatGPT Work, Word, Excel and PowerPoint — not ordinary chat — and that included usage is consumed first, with overflow drawing on the credit balance. Two terms are worth noting before you treat credits as a top-up: they are "valid for 12 months from purchase", after which unused credits expire and do not roll over, and they are "non-refundable unless required by law or contract". OpenAI publishes credit rates per million tokens for those features but no dollar price per credit, and says prices "can vary by account, region, and plan", so the cash cost of a credit cannot be stated here. OpenAI also states explicitly that these are not API credits. Checked September 21, 2026.
Can I move my ChatGPT history into TypingMind, or the other way?
One direction is documented and the other is not, which is an asymmetry worth planning around. TypingMind documents an import path for a ChatGPT Plus export under Settings → App data & storage, alongside its own Export/Import that produces JSON it says "might include chats, prompts, plugins, folders, AI Agents, model settings, and more". Going the other way, a documentation search across LibreChat, Open WebUI, Cherry Studio, Chatbox, Jan, Msty Studio and ChatWise on September 19, 2026 found none of them documenting an importer for TypingMind's export format — absence from documentation is not proof no route exists, but nobody is supporting one. Treat an export as a backup and a source you copy prompts out of by hand, and keep the old tool available while you trial the new one.
Checked September 21, 2026 for this page: TypingMind's buy page, Teams pricing page, changelog, GitHub licence file and its install, proxy, MCP, cloud-sync, cloud-storage, export/import, custom-model and Teams-comparison documentation, plus OpenAI's published API pricing, each fetched directly; and the four CORS preflights behind the table above — two Kunavo routes, OpenAI and Anthropic — run from a terminal. Carried from research checked the same day and not re-fetched here: OpenAI's help-centre articles for Plus, Go, Business, Pro and usage credits, and the ChatGPT plan comparison table — help.openai.com and chatgpt.com both refuse or geo-redirect requests from this network. Not verified at all, and therefore not stated: the current US price of ChatGPT Go or Pro, the dollar value of a ChatGPT credit, TypingMind's cloud-storage prices, whether a custom endpoint works on an unlicensed TypingMind install, whether TypingMind Teams routes custom endpoints server-side, and what TypingMind's Setapp Mac listing does with a custom endpoint. Kunavo token rates come from the live catalog, and every dollar example here is illustrative token arithmetic rather than a measured bill.