MLX, LiteRT-LM, and CoreML benchmarked for iPhone LLM performance

A recent benchmark tested four on-device LLM runtimes on an iPhone 17 Pro, comparing decode speed and memory usage. MLX emerged as the fastest for general-purpose models like Qwen 3.5 2B, while LiteRT-LM excelled specifically with Gemma 4 E2B. For memory-constrained scenarios, CoreML with the Apple Neural Engine offered significant advantages, using substantially less RAM. AI

IMPACT Provides crucial performance data for developers choosing on-device LLM runtimes for iPhones, impacting app efficiency and user experience.

RANK_REASON Benchmarking study comparing multiple software runtimes for a specific hardware platform. [lever_c_demoted from research: ic=1 ai=0.7]

Read on dev.to — LLM tag →

• Apple Neural Engine
• Gemma 4 E2B
• iPhone 17 Pro
• LiteRT-LM
• llama.cpp
• MLX
• Qwen 3.5 2B

• product
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →