Quickstart

Android: install TokForge from Google Play like any other app, or sideload the signed APK from Discord.

iPhone and iPad: install Apple TestFlight, then open the TokForge TestFlight invite and tap Install. The screenshots below are from Android; the iPhone flow mirrors them.

# Optional: sideload the APK on Android
adb install tokforge-beta.apk

On first launch, you'll see the welcome screen. Choose your preferences:

• Memory Mode: Select normal or extended memory (enables background memory extraction for continuity across conversations)
• Thinking Toggle: Enable or disable thinking/reasoning mode support

Tap Get Started to begin onboarding.

TokForge automatically profiles your device, detecting your SoC, RAM, GPU capabilities, and backend support. Based on your device tier, we recommend a model optimized for your hardware:

• 4GB RAM: Qwen3-0.6B-MNN (~0.35GB), the fastest option
• 6-8GB RAM: Qwen3-1.7B-MNN (~1.0GB) or Qwen3.5-4B-MNN (~2.3GB)
• 12GB RAM: Qwen3-8B-MNN (~4.8GB) or Qwen3.5-9B-MNN (~5.4GB)
• 16GB+ RAM: Qwen3-14B-MNN (~8.5GB) for maximum quality

The app automatically selects the best backend: MNN (GPU-accelerated on OpenCL/Vulkan) or GGUF (CPU fallback). No manual configuration needed.

TokForge downloads models directly from Hugging Face with intelligent retry and real-time progress tracking.

You'll see:

• Live download speed and ETA
• Auto-retry with exponential backoff on network hiccups
• Disk space checks to prevent corruption

Once complete, the model is cached locally and ready to use. You can download additional models anytime from the Models tab.

TokForge offers an optional speculative decoding acceleration pack: a lightweight draft model that runs alongside your main model to speed up generation.

On supported devices, speculative decoding pairs a small draft model with your main model to produce accepted tokens faster, with identical output. How much it helps depends on the device, the model and the kind of text, and the app disables it where it does not help. The draft model (typically 0.6B) is downloaded separately and toggled in Settings.

You can skip this and enable it later in Settings → Advanced.

You're all set! Open a chat and start conversing. TokForge provides:

• Streaming tokens: watch the model generate in real time with live tok/s counter
• Kokoro TTS: high-quality offline text-to-speech with 11 voices (Settings → Voice)
• Per-conversation settings: adjust temperature, sampling, and other parameters per character
• Markdown rendering and code syntax highlighting
• Thinking mode: collapsible <think> blocks show reasoning in real time
• Background memory: automatically extracts and stores facts across conversations for continuity

All conversations run 100% on-device. Nothing leaves your phone.