LemoLM

Explore local AI on your iPhone or iPad.

LemoLM is a local-first LLM playground for testing and comparing on-device language models. Try different local AI engines, compare model behaviour, and see what compact language models can do directly on your device.

The free version includes:

• Mock Engine for safe testing
• Apple Foundation Models where available
• SmolLM2 1.7B as a free downloadable GGUF model

The optional Full Model Playground Unlock enables access to additional supported local model options, including models for lightweight testing, daily use, reasoning, coding, and larger benchmark comparisons.

Supported downloadable models may include options such as Qwen, SmolLM, DeepSeek, Phi, Sakana, and Mistral-family GGUF models, depending on availability, licensing, and device capability.

LemoLM helps you:

• Test local model responses • Export conversations as markdown files • Compare different model behaviours
• Add personal star ratings and notes
• View model size, runtime, licence, and source information
• Use Load Diagnostics to troubleshoot local model loading • Use Performance Snapshot to detail recent information on generations, including timing, token estimates, speed, memory readings, and memory warning status. • Explore local AI without relying on a remote server • Models run on device keeping conversations private

LemoLM is local-first. Conversations are stored on your device, and supported local engines process prompts and responses on-device. Model downloads require internet access and may connect to third-party model hosts such as Hugging Face. Once installed, supported downloadable models run locally.

Performance depends on your device, available memory, iOS/iPadOS version, thermal state, and the size of the selected model. Compact models are intended for broader device testing, while larger downloadable GGUF models are best suited to newer high-memory iPhone and iPad models. For iPad, M-series devices are recommended for larger local models.

On iPhone, newer Pro-class devices are recommended for larger models, while compact models are better suited to older or lower-memory devices.

Please note: compact on-device language models can hallucinate, misunderstand prompts, omit context, or provide inaccurate information. LemoLM is intended for experimentation, learning, and model comparison. Always verify important information.