Can I run it? / iPhone
Best LLMs to run on an iPhone
Updated November 2026
As of November 2026, the best model to run on an iPhone is Nanbeige4.2 3B by Nanbeige (Kanzhun) (Small LM Index 67.2), at about 9 tokens/s with 4-bit weights. For faster replies, LFM3 1.7B (61.2) runs at about 23 tokens/s.
Recent iPhones have 8 GB of memory (12 GB on the 17 Pro), but iOS lets one app use only about half of it, so models up to about 4 GB fit. About 4.0 GB is usable for a model, at 60 GB/s. The figures below assume 4-bit weights and a 4K-token context, and only include models that generate at least 5 tokens per second.
Models that run well on an iPhone
The 25 best of the 29 models that run at 5+ tokens/s, ranked by the Small LM Index.
| # | Model | Index | Memory | Speed |
|---|---|---|---|---|
| 1 | Nanbeige4.2 3B Nanbeige (Kanzhun) · 4-bit · Apache-2.0 | 67.2 | 3.5 GB | ≈ 9 tok/s |
| 2 | Spark-X2.6 4B SparkLLM (XHToken) · 4-bit · Apache-2.0 | 66.5 | 3.3 GB | ≈ 10 tok/s |
| 3 | Spark-X2.5 4B SparkLLM (XHToken) · 4-bit · Apache-2.0 | 65.9 | 3.3 GB | ≈ 10 tok/s |
| 4 | LFM3 1.7B Liquid AI · 4-bit · LFM Open License | 61.2 | 1.6 GB | ≈ 23 tok/s |
| 5 | MiniCPM5 2B OpenBMB (MiniCPM) · 4-bit · Apache-2.0 | 60.9 | 2.3 GB | ≈ 16 tok/s |
| 6 | Qwen3 4B Thinking 2507 Alibaba (Qwen) · 4-bit · Apache-2.0 | 59.8 | 3.6 GB | ≈ 10 tok/s |
| 7 | Qwen3.5 4B Alibaba (Qwen) · 4-bit · Apache-2.0 | 59.7 | 3.1 GB | ≈ 10 tok/s |
| 8 | Nemotron Labs Dory 1.0 4B NVIDIA · 4-bit · OpenMDW-1.1 | 58.6 | 3.3 GB | ≈ 10 tok/s |
| 9 | Granite 4.2 3B IBM (Granite) · 4-bit · Apache-2.0 | 57.3 | 2.7 GB | ≈ 13 tok/s |
| 10 | Ministral 3 3B Mistral AI · 4-bit · Apache-2.0 | 53.8 | 2.8 GB | ≈ 13 tok/s |
| 11 | Spark-X2.6 1.7B SparkLLM (XHToken) · 4-bit · Apache-2.0 | 52.3 | 1.7 GB | ≈ 23 tok/s |
| 12 | LFM2.5 2.6B Liquid AI · 4-bit · LFM Open License | 44.7 | 2.3 GB | ≈ 15 tok/s |
| 13 | EXAONE 4.0 1.2B LG AI Research (EXAONE) · 4-bit · EXAONE License (non-commercial) | 42.5 | 1.6 GB | ≈ 31 tok/s |
| 14 | Gemma 4 E2B Google · 4-bit · Apache-2.0 | 37.4 | 3.7 GB | ≈ 17 tok/s |
| 15 | SmolLM3 3B Hugging Face (SmolLM) · 4-bit · Apache-2.0 | 36.8 | 2.7 GB | ≈ 13 tok/s |
| 16 | Qwen3.5 2B Alibaba (Qwen) · 4-bit · Apache-2.0 | 35.9 | 2.0 GB | ≈ 17 tok/s |
| 17 | LFM2.5 1.2B Thinking Liquid AI · 4-bit · LFM Open License | 34.7 | 1.3 GB | ≈ 34 tok/s |
| 18 | MiniCPM5 1B OpenBMB (MiniCPM) · 4-bit · Apache-2.0 | 29.4 | 1.3 GB | ≈ 37 tok/s |
| 19 | LFM2.5 1.2B Instruct Liquid AI · 4-bit · LFM Open License | 25.4 | 1.3 GB | ≈ 34 tok/s |
| 20 | MiniCPM-V 4.6 OpenBMB (MiniCPM) · 4-bit · Apache-2.0 | 24.7 | 1.4 GB | ≈ 30 tok/s |
| 21 | Gemma 3n E2B Google · 4-bit · Gemma | 20.9 | 4.0 GB | ≈ 20 tok/s |
| 22 | Gemma 3 4B Google · 4-bit · Gemma | 19.4 | 3.4 GB | ≈ 9 tok/s |
| 23 | Llama 3.2 3B Meta · 4-bit · Llama 3.2 Community | 19.1 | 3.0 GB | ≈ 12 tok/s |
| 24 | LFM2.5 350M Liquid AI · 4-bit · LFM Open License | 16.6 | 0.8 GB | ≈ 113 tok/s |
| 25 | Phi-4-mini Microsoft (Phi) · 4-bit · MIT | 15.8 | 3.4 GB | ≈ 10 tok/s |
Change the context length or quantization, or see models that only just fit, in Can I run it? for this device.
How to run Nanbeige4.2 3B on an iPhone
- Install an app. PocketPal AI (free, open source) downloads GGUF models from Hugging Face; Google AI Edge Gallery runs Gemma and other LiteRT models.
- Load Nanbeige4.2 3B. In PocketPal, search Hugging Face for bartowski/Nanbeige_Nanbeige4.2-3B-GGUF and pick the Q4_K_M file.
- Build it into an app. For developers: MLX Swift or llama.cpp run these models inside an iOS app.
Every model page lists the same commands for its own files. Nanbeige4.2 3B's page also has a memory table for other context lengths.
Questions
What is the best LLM for an iPhone?
As of November 2026, the best model to run on an iPhone is Nanbeige4.2 3B by Nanbeige (Kanzhun) (Small LM Index 67.2), at about 9 tokens/s with 4-bit weights. For faster replies, LFM3 1.7B (61.2) runs at about 23 tokens/s.
What is the biggest model that runs on an iPhone?
Gemma 3n E2B (5.4B parameters) fits at 4-bit, needing 4.0 GB and generating about 20 tokens/s. Bigger models may fit at lower quality or run more slowly; try them in the Can I run it? tool.
How much memory can a model use on an iPhone?
About 4.0 GB of the 8 GB. Recent iPhones have 8 GB of memory (12 GB on the 17 Pro), but iOS lets one app use only about half of it, so models up to about 4 GB fit.
How accurate are these speeds?
They are estimates from memory bandwidth (60 GB/s here), which limits how fast a model generates text, with a typical efficiency for this kind of hardware. Expect real speeds within about a third either way. Long prompts take longer to read before the first word appears.