Token: Local AI Studio

Token: Local AI Studio

Local AI, chat, voice & images

2 ratings
2 reviews
Free

Details

Released
May 30, 2026
Updated
November 9, 2026

Features

Token: Local AI Studio video #1 for iPhone
Token: Local AI Studio screenshot #2 for iPhone
Token: Local AI Studio screenshot #3 for iPhone
Token: Local AI Studio screenshot #4 for iPhone
Token: Local AI Studio screenshot #5 for iPhone
Token: Local AI Studio screenshot #6 for iPhone
Token: Local AI Studio screenshot #7 for iPhone
iphone
iphoneDuo
🖼️Get Icon
Icons↘︎

About

This app provides a private AI studio on your iPhone, allowing you to download and run various open AI models locally. Engage in conversations, generate images, and utilize tools like web search and document analysis, all while keeping your data on your device.

✓Download and run open AI models locally
✓On-device chat, reasoning, and coding
✓Image generation with local models
✓Voice conversations with on-device speech recognition
✓Private by design, no account or analytics
✓Local API server for compatible tools
Show original description
+2

What's New in Token

2.1.6

November 9, 2026

• Voice chat is steadier. Fixed a freeze on iOS 27 that could close the app in the middle of a voice conversation. • Spoken replies no longer go silent when Apple's voice stalls or your audio switches to AirPods: voice mode recovers and the microphone turns back on by itself. • Voice chat answers directly and briefly, without hidden reasoning, so replies start sooner and the phone stays cooler. • Leaving the app pauses voice mode. Tap Try again to continue the same conversation. • If a voice reply fails, a message now tells you why. • When a downloaded voice doesn't fit in memory next to the chat model, Token AI speaks with the Apple voice instead. • Fixed: voice chat failed with "Error reading 'config.json'" when the chat model was a GGUF imported from Files.

More

User reviews

Voice is broken

On an iPhone 17 using a 3B model tried a 4B model Q2 and Q4 first so I switched to a 3B model that was even lighter only two gigs and it would use Voice for like a few sentences back-and-forth and then the voice would just stop or the app just crashed and I have to reboot it. I’ve actually had to restart my phone so do not recommend it’s free but there’s tons of these apps now I wish they would just stop making them and make a good one. I mean, I would pay like seven bucks for one that actually works.
Show more

Response from developer

Thanks for the detailed report, and sorry voice mode let you down. Your description helped me find problems in the voice path on iOS 27 that cause speech to stop after a few turns and can freeze the app. Imported GGUF models also failed to load in voice mode. Fixes for both are coming in the next update. I'd love for you to try it again then.

I want to like this app BUT...

Each time I send a message it crashes and it's very frustrating

Response from developer

Hi Ericmst, thank you for reporting this, and I’m sorry for the frustration. I've fixed a crash that could occur when sending messages at some device models, and the update is now in testing. I'm working to make it available soon. Thank you for your patience and for giving SiteAgent a chance.

Developer apps

FAQ

What AI models can I use with Token?

Token supports a variety of models including experimental ones like Edge0 35B and MiniCPM-V 4.7 35B, as well as Ternary Bonsai 2 9B, Bonsai 27B, Qwen3.5, Qwen3, Llama 3, Gemma, Phi-3.5, LFM2.5, Nemotron, and Granite. You can also import compatible MLX or GGUF models from Files.

Does Token have ads?

No, Token is private by design and does not contain advertisements. Conversations are stored only on your iPhone, and network usage for features like web search is optional.

How much storage do model downloads require?

Model files are downloaded separately and can require several gigabytes of storage. The specific memory requirements for models vary, with some needing up to 12 GB of memory on your iPhone.

Can I use Token offline?

Yes, Token is designed to run AI models locally on your device. This means you can engage in chat, reasoning, and image generation even without an active internet connection, though features like web search require network access.

What is the age rating for Token?

Token has an age rating of 17+, indicating it is intended for mature users. This is likely due to the advanced nature of AI technology and the potential for complex outputs.

How often is Token updated?

The latest version of Token is 2.1.6, which was last updated on November 9, 2026. This indicates a relatively frequent update cycle for the application.

Can I integrate Token with other apps or services?

Token offers an authenticated local API server compatible with OpenAI, Anthropic, and Ollama tools. It also includes Shortcuts and Siri actions, and a share extension for enhanced integration possibilities.

What are the system requirements for Token?

Token requires iOS 27 or later and works on iPhone Duo devices. The availability of specific AI models depends on your iPhone's memory capacity.