MorseSign inStart freeOpen Morse

Morse vs NanoGPT: pay-per-prompt privacy or a privacy mode per chat

The Morse teamPublished Facts checked 9 min read

Morse vs NanoGPT comes down to how you'd rather pay and who sets the privacy of each prompt. As of November 2026, NanoGPT lets you chat with 1,263 models with no account, bills per prompt from a prepaid balance, keeps history in your browser, and offers four privacy levels per request, from standard routing, where the provider may keep your prompt, up to a browser-encrypted Private Mode. Morse is a subscription with 80+ models, a privacy mode per chat, end-to-end encryption your browser verifies, and history encrypted on your device. Choose NanoGPT if you want pay-as-you-go with no sign-up; choose Morse if you want to pick the privacy of each chat yourself.

We read NanoGPT's privacy explainer, privacy policy and terms (both last updated November 19, 2026), its pricing page, models page, privacy guide, Private Mode verification page, and two of its FAQ answers, all on November 19, 2026. We didn't open an account or send prompts. NanoGPT here is the chat service run by NanoGPT LLC, not Andrej Karpathy's nanoGPT, a training codebase whose README now calls it "very old and deprecated." NanoGPT earns credit for candor: its own explainer has a table of who can read your prompt at each level, and says plainly where it has to take a provider's word.

At a glance

As of November 19, 2026. Every NanoGPT fact links to NanoGPT's own page.

MorseNanoGPT
Account neededAn emailed code or Google sign-in; a first message from the home page opens a guest chat without oneNone: "you can use NanoGPT with a generated user ID and avoid sharing an email" (NanoGPT)
Where your prompt is readE2EE: only inside a secure enclave your browser verified. Private: at a host under a zero data retention agreement. Anonymous: at the provider, without knowing it's youBy default, by NanoGPT in transit and by the provider, "which may keep prompts under its own terms"; in Private Mode, only by code in an attested enclave (NanoGPT)
Who verifies itIn E2EE, your browser checks the enclave's hardware attestation before anything is sentIn Private Mode, your browser or a local proxy; other TEE routes vary "by provider, transport, attestation" (NanoGPT)
Where chat history livesIn your browser, encrypted, not on Morse's serversIn your browser by default, with opt-in cloud sync (NanoGPT)
TrainingMorse doesn't train AI on your chatsNanoGPT doesn't; on standard routing, "A few routes allow training" (NanoGPT)
AdsNoneDoesn't use your content for advertising, and "no longer loads advertising pixels" (NanoGPT)
Models80+, open and closed1,263 text, image, video, 3D, audio, and embedding models (NanoGPT)
Free15 messages a day, with the model picked for you"One free model, available on web only" (NanoGPT)
Paid$19 or $49 a month ($17 or $44 billed yearly)Pay as you go at list prices, or Pro at $12 a month for open-source models (NanoGPT)

Is NanoGPT private?

From NanoGPT itself, mostly yes. Its policy says it does "NOT store prompt or conversation content on our servers" by default, and doesn't train on your input or output or sell your content (NanoGPT). From the model provider, it depends on the level you pick. NanoGPT's explainer lists four, strongest first (NanoGPT):

NanoGPT levelWho can read your prompt, in NanoGPT's wordsWhat it rests on
Private ModeCode inside attested enclaves; NanoGPT sees "account, route, timing, size and usage, not the content"Attestation you can re-check, which "doesn't prove which model answered or what is retained"
TEE models"NanoGPT handles the request in transit"; hardware isolation protects it from the hostAttestation; "Coverage and verification vary by route"
ZDR onlyNanoGPT in transit, and the provider while it processes the request"The provider's statement. We can't inspect its servers"
Standard routing (the default)NanoGPT in transit, and the provider, "which may keep prompts under its own terms""Each provider's terms"

Standard routing, the weakest level, is the default. NanoGPT calls its zero data retention "best effort," says it "does not run model inference itself," and names OpenRouter among the intermediaries that screen every request for child sexual abuse material and may keep flagged content (NanoGPT). It also keeps "limited non-content metadata" for billing, reliability, and abuse prevention. Its policy names dozens of hosts, each with its own retention note, some of which don't guarantee zero retention at all (NanoGPT). What a zero retention promise does and doesn't cover is in what is zero data retention.

Do you need an account or email for NanoGPT?

No. NanoGPT's FAQ says you can use it "without creating an account or providing an email address" and that it needs no KYC or ID check, though card and crypto payment providers may run their own checks (NanoGPT). To move a balance to another device, you generate a revocable sign-in token in Settings, or make an account "with any anonymous email" (NanoGPT). Its FAQ lists 29 crypto options, Monero and Bitcoin among them, plus cards through Stripe, and gives every crypto deposit a 3% bonus (NanoGPT). NanoGPT is ahead of Morse here, since you can reach paid models with no account at all.

Does NanoGPT tell model providers who you are?

Not by name, but often by a stable hashed ID. NanoGPT says it normally doesn't send your name, email, or NanoGPT user ID to providers, and doesn't pass your IP address to them directly (NanoGPT). Its policy adds that "When you use OpenAI, Meta, Anthropic or Google Gemini models, most routes send the model provider a stable hashed account or session identifier," which "can associate requests from the same account or session" (NanoGPT). Zero data retention routing and redaction don't turn it off. A separate setting, Skip routes that need a stable account ID, does for text requests, at the cost of losing models that only run on those routes; image, audio, video, embedding, and batch requests aren't covered yet.

What is NanoGPT's Private Mode, and how is it checked?

It's NanoGPT's strongest level, and you can verify it. Your browser or local proxy checks the enclave before NanoGPT receives the request, evidence counts as fresh for five minutes, and failed or mismatched evidence "fails closed" with no fallback to plaintext (NanoGPT). Each reply carries a receipt you can re-check with a published command-line verifier. NanoGPT lists the limits itself. Attestation covers the enclave operator's router, "not which model answered," receipts are "browser-generated and unsigned," and Private Mode has no web search, link reading, or image generation (NanoGPT). For one provider's standard TEE route, it says it doesn't yet validate the GPU evidence against NVIDIA's attestation service (NanoGPT). What a browser's enclave check proves in general is in how to verify an AI enclave's attestation.

Where does NanoGPT keep your chat history?

In your browser, unless you turn on sync. NanoGPT's Conversation Sync recommends passkey or passphrase encryption, which encrypts chats in your browser before upload; its "Convenience mode" lets NanoGPT's infrastructure decrypt them (NanoGPT). Two features can send chat text elsewhere. If you use Memory, its default suggestion analyzer sends excerpts through NanoGPT to a TEE-backed provider even when sync is off. Context Memory, a long-chat feature run by PolyChat and off by default, keeps a memory of the conversation for 30 days by default, and NanoGPT says that for "very long chats on some subscription-covered models, we may use Context Memory ourselves without you turning it on." NanoGPT says this doesn't happen for ZDR-only or EU-only requests, in Private Mode, or with your own key (NanoGPT).

Does NanoGPT have uncensored models?

Yes, many, billed per token. Its pricing page lists GLM 5.3 Flash Uncensored at $0.20 per million input tokens and abliterated DeepSeek V4.1 Flash versions, and notes that Pro leaves out "Some uncensored models, such as GLM 5.3 Uncensored" (NanoGPT). On Morse, uncensored models come with Pro and Max: DeepSeek V4.1 Flash Uncensored is a Pro text model, and picking one gets answers other apps refuse, such as dark fiction, straight medical and legal answers, and help testing your own security.

How much does NanoGPT cost compared with Morse?

NanoGPT charges per token, Morse per month. NanoGPT bills text "at list prices" with "no percentage on top" and no deposit fees, from $0.10 in crypto or $1 by card; pinning a provider adds 5%, and bring-your-own-key requests cost 5% of normal (NanoGPT). Its Pro plan, $12 a month, includes 60 million input tokens a week on open-source models like GLM, DeepSeek, and Kimi, but not "Most closed models, such as Claude, GPT, Gemini and Grok" (NanoGPT). Its terms add a $20 fee on any card chargeback (NanoGPT).

Take NanoGPT's DeepSeek V4.1 Flash Uncensored, listed at $0.33 per million input tokens and $1.32 per million output tokens on November 19, 2026 (NanoGPT). A 1,000-token reply to a 2,000-token prompt costs about $0.002, so $19 covers roughly 9,600 replies that size. On Morse Pro, DeepSeek V4.1 Flash Uncensored is one of the Pro models with unlimited messages for $19 a month, or $17 billed yearly, along with E2EE chats, memory across chats, and deep research in Anonymous mode. Max is $49 a month, or $44 billed yearly, with 5,500 credits.

What does NanoGPT do better than Morse?

On its own pages, read November 19, 2026, NanoGPT leads in five places:

  • No account, no subscription. You can top up from $0.10 and use paid models with a generated user ID (NanoGPT).
  • Breadth. 1,263 models, with image, video, 3D, and audio among them (NanoGPT), against Morse's 80+.
  • Prices you can audit. Every chat completion and Responses API response "includes the exact cost charged" (NanoGPT).
  • Built-in redaction. Optional PII redaction swaps names, emails, and numbers for placeholders before the provider sees them (NanoGPT).
  • A developer API. An OpenAI-compatible API that works in tools like Claude Code and Cursor (NanoGPT). Morse also has a pay-as-you-go private inference API.

How Morse keeps your prompts private without route settings

On NanoGPT, how private a prompt is depends on its route: which of four levels it takes, which host serves it, and which settings, IDs, and helpers travel with it. On Morse you pick a privacy mode for each chat, and that mode decides who can read what you send.

  • E2EE, for what you'd send in NanoGPT's Private Mode. Your browser checks the enclave's hardware attestation, seals the message for that enclave alone, and Morse's servers relay text they can't read. Each reply has a lock you can open to see every check that passed. The enclaves are run by third-party providers, not by Morse, and the guarantee rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models with no web search or image generation, and comes with Pro.
  • Private, the default, for open models and Grok without a route to choose. DeepSeek, Kimi, GLM, and xAI's Grok run here. Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing, and where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
  • Anonymous, for GPT, Claude, and Gemini under Morse's account. Our proxy removes your account, email, and IP address and sends the message under Morse's own account, so the provider sees it without knowing it's you. It may keep it for a limited time to check for abuse, under its own policy, so leave out details that point to you.

Your chat history is stored in your browser, encrypted, not on Morse's servers. Morse doesn't train AI on your chats, sell your data, or show ads, and because we don't store chats, we usually have very little to give authorities. Clearing your browser deletes those chats, and Morse can't recover them, so keep a copy of anything you need.

NanoGPT or Morse: which should you choose?

  • Choose NanoGPT if you want to pay per prompt with no account or email, reach 1,263 models including video and 3D, redact personal data automatically, and you're happy to pick a privacy level, a route, and the ID setting for what you send.
  • Choose Morse if you'd rather pay one monthly price, pick a privacy mode for each chat, have your browser verify the enclave in E2EE, use uncensored models on Pro, and keep your history encrypted on your device.

Send a first message from the Morse home page without an account, or start free with 15 messages a day.

Morse publishes this comparison. We've tried to keep it fair; if something's wrong or out of date, email hello@trymorse.ai.

NanoGPT is a trademark of NanoGPT LLC. Morse is not affiliated with or endorsed by NanoGPT LLC.

Common questions

Is NanoGPT the same as Karpathy's nanoGPT?

No. nanoGPT on GitHub is Andrej Karpathy's code for training small GPT models, which its README now calls 'very old and deprecated.' NanoGPT at nano-gpt.com is a pay-as-you-go AI chat service run by NanoGPT LLC.

Does NanoGPT keep my prompts?

NanoGPT says it doesn't log or retain prompt and response content during normal inference, though it keeps billing and security metadata. What the model provider keeps depends on the route: on standard routing, the default, the provider may keep prompts under its own terms.

Can I turn off the hashed account ID NanoGPT sends to providers?

Yes. NanoGPT's privacy policy says the setting 'Skip routes that need a stable account ID' stops it for text requests, though models that run only on those routes become unavailable, and image, audio, and video requests aren't covered yet.

Is NanoGPT cheaper than Morse?

For light use, usually. NanoGPT bills each request at the model's list price from a top-up as small as $0.10 in crypto or $1 by card. Morse Pro is a flat $19 a month, or $17 billed yearly, with unlimited messages on Pro models such as DeepSeek V4.1 Flash Uncensored.

Ask what you actually want to ask.

80+ models, characters, and images, with every chat encrypted and kept on your device. Free to start.

Start a private chat

Read next