Device Compatibility

Community-tested devices. If you've run llamdrop on your device, open a PR to add it here.

---

How to Read This Table

Column	Meaning
Device	The actual device name
Chipset	The processor inside
RAM	Total device RAM
Platform	How llamdrop was run
Max Stable Model	Largest model confirmed to run without crashing
Vulkan	Whether GPU acceleration worked
Notes	Anything specific to this device

---

✅ Confirmed Working

Device	Chipset	RAM	Platform	Max Stable Model	Vulkan	Notes
Oppo F19 Pro+ 5G	Dimensity 800U	8GB	Android / Termux	Qwen2.5-1.5B Q4	❓	Original llamdrop dev device

---

🔄 Reported Working (Not Yet Verified)

Device	Chipset	RAM	Platform	Reported By
Samsung Galaxy S25	Snapdragon 8 Elite	12GB	Android / Termux	Community

---

Platform Support Status

Platform	Status	Architecture	Notes
Android via Termux	🎯 Primary test platform	ARM64	Built and tested here first. GPU intentionally CPU-only (Android GPU slower than CPU for inference).
Linux laptop / desktop	✅ Fully supported	x86_64 or ARM64	Any distro. Vulkan/CUDA/ROCm auto-detected.
Raspberry Pi 4 / 5	✅ Fully supported	ARM64
Chromebook (Linux mode)	🔄 Should work	ARM64 or x86_64
Orange Pi / other SBCs	🔄 Should work	ARM64	Any ARM64 Linux
iOS	❌ Not supported	—	No proper terminal environment
macOS	❌ Dropped in v0.10.0	—	Desktop/Workstation scope removed
Windows	❌ Dropped in v0.10.0	—	Desktop/Workstation scope removed

---

Vulkan GPU Acceleration

llamdrop auto-detects GPU support at launch. The decision is shown in the Device Profile on first launch and in the Device Info screen.

GPU Type	Devices	Status
Adreno (Qualcomm)	Snapdragon phones	⚠️ Detected but disabled on Android — driver present, hardware not usable for AI inference
Mali (ARM)	MediaTek, some Samsung	⚠️ Detected but disabled on Android — slower than CPU for inference
Desktop Vulkan	Linux laptops/desktops with AMD or Intel	✅ Active
CUDA	NVIDIA desktop/laptop GPUs	✅ Active — auto-selected at install time
ROCm	AMD desktop GPUs (Linux)	✅ Active — auto-selected at install time

Why is Android GPU disabled? Android Vulkan drivers for Mali and Adreno are designed for graphics, not general compute. Running AI inference through them is typically slower than CPU and can cause crashes. llamdrop makes this decision automatically.

GPU layers are offloaded conservatively on desktop Vulkan based on available RAM:

• 4GB+ free → 20 layers
• 2.5GB+ free → 10 layers
• 2GB+ free → 5 layers
• Under 2GB → CPU only

IQ quants (IQ2/IQ3/IQ4) always run CPU-only regardless of GPU availability — they are incompatible with the Vulkan compute path.

---

Device Tiers and What They Mean

llamdrop classifies every device into one of 5 tiers at launch. Your tier determines which models appear in the browser. Note: We intentionally drop support for hardware tiers above High (16GB+) to focus exclusively on accessible hardware.

Tier	Available RAM	Typical devices
Micro	< 1 GB	Very old phones, minimal Linux
Low	1 – 3 GB	Budget Android phones, Raspberry Pi 4 (4GB)
Low-Mid	3 – 6 GB	Mid-range phones, older laptops
Mid	6 – 12 GB	Modern phones, mainstream laptops
High	12 – 16 GB+	Flagship phones, gaming laptops

Hardware Restrictions & Overrides: llamdrop enforces safety limits (like thermal throttling and battery thresholds) because mobile hardware isn't designed for sustained 100% CPU loads. While we enforce these out of the box to prevent damage or crashes, power users can override these limits (e.g. allow_thermal_melt) using config.json at their own risk.

---

Devices We Need Testers For

Android phones — Budget (4–6GB RAM)

• Helio G series (Redmi Note, Narzo, Poco M series)
• Snapdragon 6xx (Galaxy A series, Moto G series)

Android phones — Mid-range (8GB RAM)

• Dimensity 8xx devices
• Snapdragon 7xx devices

Android phones — Flagship (12GB+ RAM)

• Snapdragon 8 Gen series
• Dimensity 9xxx series

Single Board Computers

• Raspberry Pi 4 (4GB and 8GB)
• Raspberry Pi 5
• Orange Pi 5
• Rock Pi

Linux Devices

• Old Intel laptops (4–8GB RAM)
• AMD budget laptops

---

Add Your Device

1. Run llamdrop successfully on your device
2. Run llamdrop doctor and note all the green checks
3. Note the largest model that runs without crashing and its t/s speed
4. Open a Pull Request editing docs/DEVICES.md

Template:

| Device Name | Chipset | X GB | Android/Termux or Linux | Model + Quant | ✅/❌/❓ | Notes |

This list is 100% community-maintained. The more devices here, the more people llamdrop can confidently help.