Everything MyBenAI does
A complete AI stack — chat, voice, vision, image generation, tools, agents, research, documents and automation — running on the phone in your hand. This page lists all of it, and says plainly which parts the free edition includes.
Two editions, one app
Same codebase, same privacy model, same offline behaviour. The difference is four features and the size of the models you can load.
MyBenAI Lite
Free
Everything below marked Free, plus a 7-day full-access trial from first launch. A one-time $2.99 unlock opens the rest, forever, with no subscription.
MyBenAI
$2 one-time
Every feature, unconditionally. No paywall screen, no in-app purchases, not even the billing permission in the manifest. The cheaper of the two routes to the full app.
| Capability | Lite (free) | MyBenAI ($2) |
|---|---|---|
| Chat with local LLMs | Up to mid-tier models | All tiers |
| Largest models (4B–9B) | Unlock | Included |
| Voice: Whisper STT, TTS, hands-free | Included | Included |
| Vision (ask about images) | Unlock | Included |
| On-device image generation | Unlock | Included |
| Deep research (cited reports) | Unlock | Included |
| Tools & function calling | Included | Included |
| Web search & page reading | Included | Included |
| Sub-agents & delegation | Included | Included |
| Documents, folders & notebooks (RAG) | Included | Included |
| Long-term memory | Included | Included |
| Personas | Included | Included |
| Scheduled jobs | Included | Included |
| MCP & REST connectors | Included | Included |
| OpenAI-compatible LAN API | Included | Included |
| Battery & thermal guard | Included | Included |
| Offline mode & network log | Included | Included |
| Ads, trackers, profiling | None | None |
| Anonymous usage stats | On by default, switchable off | On by default, switchable off |
| Account required | No | No |
On iPhone and iPad there is a single free app with the same one-time unlock — the App Store does not allow the same app under two listings.
The app itself
Real screenshots from an Android phone and an iPhone — not mockups, and not a render of a phone with a picture in it.
Every feature, explained
One page each — what it does, how it works on a phone, and where the limits are.
Conversation
Offline chat Free
Full multi-turn conversation with a local LLM running in your phone's own memory. No account, no API key, and identical behaviour in airplane mode.
Automatic model selection Free
MyBenAI profiles your RAM, cores and thermal headroom on first launch and downloads the best models your phone can actually run. No GGUF hunting.
Model catalog Free
Every model is a catalog entry refreshed over the air, so a better model for your phone arrives without an app update — and you can install or delete any of them by hand.
Larger models Pro
The high and flagship tiers — Qwen3.5 4B, Gemma 3 4B, Phi-4 Mini, and 8–9B models on 12 GB phones. Sharper reasoning and much better tool use than the free tiers.
Personas Free
Named system prompts with their own temperature and tool allowlist, switchable per conversation — a terse coding assistant, a patient tutor, a journalling companion.
Benchmark Free
Prompt-processing and generation speed, memory headroom and thermal behaviour, per model, on your hardware — and the results feed back into the app's own estimates.
Conversation search Free
Every message is indexed on-device with SQLite full-text search, so a conversation from four months ago is one query away — and the index never leaves your phone.
Voice
Speech-to-text Free
OpenAI's Whisper running locally through whisper.cpp. Dictate, transcribe a voice memo or a meeting — no upload, no minute quota, no per-hour fee.
Hands-free mode Free
On-device voice activity detection hears where your speech starts and stops, so you talk, it answers, and you talk again without touching the screen.
Read aloud Free
Have any answer spoken back with the system voices your phone already has — useful while driving or cooking, and for anyone who would rather listen than read.
Images
Vision Pro
A local vision-language model reads receipts, whiteboards, error dialogs, forms and handwriting, and answers in the same conversation. The image never leaves your phone.
Image generation Pro
Stable Diffusion running on the phone — stable-diffusion.cpp on Android, Core ML on iPhone. Unlimited generations, no credits, no queue, nothing uploaded.
Tools
Tools & function calling Free
Calculator, calendar, contacts, weather, units, currency, dictionary, device status and more — each declaring whether it needs the network, each toggleable.
Web search Free
Search the open web, read a page as clean text, or crawl a site a few hops deep — the only features here that use the network, and offline mode switches them off.
Agents & research
Sub-agents Free
Five specialists ship with the app — Researcher, Fact Checker, Summarizer, Document Analyst, Extractor — searching and reading in parallel over one model context.
Custom agents Free
Agent definitions are data, not code: a prompt, a tool allowlist, a step budget, and an output schema that grammar-constrained decoding forces the model to match.
Deep research Pro
Plan, search, read, notice what is missing, go back for it, and write a report with numbered citations you can open. Every run is saved with its sources.
Documents & memory
Documents & RAG Free
Import PDFs, text and Markdown. They are chunked, embedded on your phone and indexed in a local vector store — contracts and medical letters never touch a website.
Folders Free
Group documents and point a conversation at one group, so a question about your thesis is not answered out of your tax paperwork.
Notebooks Free
Collect sources, keep notes beside them, and ask questions against just that collection — study guides, case files, a literature review, trip planning.
Long-term memory Free
Facts are saved only when you confirm them, and every one is listed, editable and deletable. You can always see exactly what the assistant thinks it knows about you.
Automation & connectivity
Scheduled jobs Free
Daily, weekly or every N minutes — a morning briefing, a weekly digest of a folder, a recurring summary. Jobs the OS never woke are caught up and marked as late.
MCP connectors Free
Point the app at a Model Context Protocol server and its tools appear beside the built-in ones — your home automation, your notes service, your own scripts.
REST connectors Free
Describe an endpoint, its parameters and its authentication — OAuth included — and the model can call it like any other tool.
LAN API server Free
Turn on a small server and your laptop uses the phone's model: /v1/models and /v1/chat/completions, streaming included. Anything that speaks the OpenAI API works.
Battery & performance
Battery & thermal guard Free
One authority decides what the phone can afford from battery, charging state and thermal status, and every heavy operation asks it first — then says so in plain words.
Model swapping Free
Text, vision and diffusion models are never resident together. Swaps are explicit and visible, and your conversation is saved before the context is released.
Privacy
Network log Free
Not a policy document — a list. Every outbound request is recorded with when, where to and why, for the last 30 days, readable inside the app.
Offline mode Free
Tools that need the internet then report themselves unavailable to the model, so it answers from what it has instead of stalling on a request that will never finish.
On-device data Free
Conversations, documents, memories and generated images live in the app's own storage. Delete the app and they are gone — there is no copy to request or worry about.
Questions about the two editions
What is the difference between MyBenAI and MyBenAI Lite?
MyBenAI is the paid app: everything is included unconditionally and there are no in-app purchases. Lite is free and includes chat, voice, tools, sub-agents, documents, memory, jobs and connectors. Four things are held back — image generation, vision, deep research and the largest model tiers — and one $2.99 unlock opens all four forever.
Which is the better deal?
The paid app, at $2, if you already know you want everything. Lite is the way in if you would rather try the whole thing first — it gives every new install 7 days of full access with no card on file.
Does the trial charge me when it ends?
No. It is not a subscription trial. There is nothing to cancel: when the seven days are up the four paywalled features lock and the rest of the app carries on working, free, indefinitely.
Which features need the internet?
Web search, deep research, weather, currency and crypto prices, connector calls, and downloading models or refreshing the catalog. Everything else — chat, voice, vision, image generation, documents, memory, jobs, the LAN API — runs with the radio off.
Which phones can run it?
4 GB of RAM runs 0.6–1.7B models comfortably, 6 GB runs 2–3B, 8 GB runs 4B models and vision, and 12 GB and up runs 8–9B. The app measures your phone and picks for you; image generation wants a mid-tier device or better.
What about iPhone?
One free app on the App Store with the same one-time unlock, because Apple does not allow the same app under two listings. The feature split is identical.
Start free, or buy the whole thing for $2
No account. No subscription. Nothing leaves your phone.